?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
push 553a83299288f61678d62ade87a3a2a5489a8ad8
[wine/hacks.git] / dlls / wined3d / arb_program_shader.c
bloba8d56331a869e8511231d20546ca3cdd86588aad
1 /*
2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
4 *
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13 *
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2.1 of the License, or (at your option) any later version.
18 *
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
23 *
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
27 */
28
29 #include "config.h"
30
31 #include <math.h>
32 #include <stdio.h>
33
34 #include "wined3d_private.h"
35
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
39
40 #define GLINFO_LOCATION (*gl_info)
41
42 /********************************************************
43 * ARB_[vertex/fragment]_program helper functions follow
44 ********************************************************/
45
46 /**
47 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
48 * When constant_list == NULL, it will load all the constants.
49 *
50 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
51 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
52 */
53 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info, GLuint target_type,
54 unsigned int max_constants, float* constants, char *dirty_consts) {
55 local_constant* lconst;
56 DWORD i, j;
57 unsigned int ret;
58
59 if (TRACE_ON(d3d_shader)) {
60 for(i = 0; i < max_constants; i++) {
61 if(!dirty_consts[i]) continue;
62 TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
63 constants[i * 4 + 0], constants[i * 4 + 1],
64 constants[i * 4 + 2], constants[i * 4 + 3]);
65 }
66 }
67 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
68 if(target_type == GL_FRAGMENT_PROGRAM_ARB &&
69 WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1) {
70 float lcl_const[4];
71 for(i = 0; i < max_constants; i++) {
72 if(!dirty_consts[i]) continue;
73 dirty_consts[i] = 0;
74
75 j = 4 * i;
76 if(constants[j + 0] > 1.0) lcl_const[0] = 1.0;
77 else if(constants[j + 0] < -1.0) lcl_const[0] = -1.0;
78 else lcl_const[0] = constants[j + 0];
79
80 if(constants[j + 1] > 1.0) lcl_const[1] = 1.0;
81 else if(constants[j + 1] < -1.0) lcl_const[1] = -1.0;
82 else lcl_const[1] = constants[j + 1];
83
84 if(constants[j + 2] > 1.0) lcl_const[2] = 1.0;
85 else if(constants[j + 2] < -1.0) lcl_const[2] = -1.0;
86 else lcl_const[2] = constants[j + 2];
87
88 if(constants[j + 3] > 1.0) lcl_const[3] = 1.0;
89 else if(constants[j + 3] < -1.0) lcl_const[3] = -1.0;
90 else lcl_const[3] = constants[j + 3];
91
92 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, lcl_const));
93 }
94 } else {
95 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS)) {
96 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
97 * or just reloading *all* constants at once
98 *
99 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
100 */
101 for(i = 0; i < max_constants; i++) {
102 if(!dirty_consts[i]) continue;
103
104 /* Find the next block of dirty constants */
105 dirty_consts[i] = 0;
106 j = i;
107 for(i++; (i < max_constants) && dirty_consts[i]; i++) {
108 dirty_consts[i] = 0;
109 }
110
111 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, j, i - j, constants + (j * 4)));
112 }
113 } else {
114 for(i = 0; i < max_constants; i++) {
115 if(dirty_consts[i]) {
116 dirty_consts[i] = 0;
117 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, i, constants + (i * 4)));
118 }
119 }
120 }
121 }
122 checkGLcall("glProgramEnvParameter4fvARB()");
123
124 /* Load immediate constants */
125 if(This->baseShader.load_local_constsF) {
126 if (TRACE_ON(d3d_shader)) {
127 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
128 GLfloat* values = (GLfloat*)lconst->value;
129 TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
130 values[0], values[1], values[2], values[3]);
131 }
132 }
133 /* Immediate constants are clamped for 1.X shaders at loading times */
134 ret = 0;
135 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
136 dirty_consts[lconst->idx] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
137 ret = max(ret, lconst->idx + 1);
138 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type, lconst->idx, (GLfloat*)lconst->value));
139 }
140 checkGLcall("glProgramEnvParameter4fvARB()");
141 return ret; /* The loaded immediate constants need reloading for the next shader */
142 } else {
143 return 0; /* No constants are dirty now */
144 }
145 }
146
147 /**
148 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
149 *
150 * We only support float constants in ARB at the moment, so don't
151 * worry about the Integers or Booleans
152 */
153 void shader_arb_load_constants(
154 IWineD3DDevice* device,
155 char usePixelShader,
156 char useVertexShader) {
157
158 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
159 IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
160 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
161 unsigned char i;
162
163 if (useVertexShader) {
164 IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
165
166 /* Load DirectX 9 float constants for vertex shader */
167 deviceImpl->highest_dirty_vs_const = shader_arb_load_constantsF(
168 vshader, gl_info, GL_VERTEX_PROGRAM_ARB,
169 deviceImpl->highest_dirty_vs_const,
170 stateBlock->vertexShaderConstantF,
171 deviceImpl->activeContext->vshader_const_dirty);
172
173 /* Upload the position fixup */
174 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB, ARB_SHADER_PRIVCONST_POS, deviceImpl->posFixup));
175 }
176
177 if (usePixelShader) {
178
179 IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
180 IWineD3DPixelShaderImpl *psi = (IWineD3DPixelShaderImpl *) pshader;
181
182 /* Load DirectX 9 float constants for pixel shader */
183 deviceImpl->highest_dirty_ps_const = shader_arb_load_constantsF(
184 pshader, gl_info, GL_FRAGMENT_PROGRAM_ARB,
185 deviceImpl->highest_dirty_ps_const,
186 stateBlock->pixelShaderConstantF,
187 deviceImpl->activeContext->pshader_const_dirty);
188
189 for(i = 0; i < psi->numbumpenvmatconsts; i++) {
190 /* The state manager takes care that this function is always called if the bump env matrix changes
191 */
192 float *data = (float *) &stateBlock->textureState[(int) psi->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
193 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->bumpenvmatconst[i].const_num, data));
194 deviceImpl->activeContext->pshader_const_dirty[psi->bumpenvmatconst[i].const_num] = 1;
195
196 if(psi->luminanceconst[i].const_num != -1) {
197 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
198 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
199 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
200 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
201 */
202 float *scale = (float *) &stateBlock->textureState[(int) psi->luminanceconst[i].texunit][WINED3DTSS_BUMPENVLSCALE];
203 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->luminanceconst[i].const_num, scale));
204 deviceImpl->activeContext->pshader_const_dirty[psi->luminanceconst[i].const_num] = 1;
205 }
206 }
207
208 if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
209 !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
210 float comparison[4];
211 float mul_low[4];
212
213 if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
214 comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
215 comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
216
217 mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
218 mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
219 } else {
220 comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
221 comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
222
223 mul_low[0] = 1.0; mul_low[1] = 1.0;
224 mul_low[2] = 1.0; mul_low[3] = 1.0;
225 }
226 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_cmp_const, comparison));
227 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, psi->srgb_low_const, mul_low));
228 checkGLcall("Load sRGB correction constants\n");
229 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_low_const] = 1;
230 deviceImpl->activeContext->pshader_const_dirty[psi->srgb_cmp_const] = 1;
231
232 }
233 }
234 }
235
236 /* Generate the variable & register declarations for the ARB_vertex_program output target */
237 void shader_generate_arb_declarations(
238 IWineD3DBaseShader *iface,
239 shader_reg_maps* reg_maps,
240 SHADER_BUFFER* buffer,
241 WineD3D_GL_Info* gl_info) {
242
243 IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
244 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
245 DWORD i, cur;
246 char pshader = shader_is_pshader_version(This->baseShader.hex_version);
247 unsigned max_constantsF = min(This->baseShader.limits.constant_float,
248 (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
249 UINT extra_constants_needed = 0;
250 local_constant* lconst;
251
252 /* Temporary Output register */
253 shader_addline(buffer, "TEMP TMP_OUT;\n");
254
255 for(i = 0; i < This->baseShader.limits.temporary; i++) {
256 if (reg_maps->temporary[i])
257 shader_addline(buffer, "TEMP R%u;\n", i);
258 }
259
260 for (i = 0; i < This->baseShader.limits.address; i++) {
261 if (reg_maps->address[i])
262 shader_addline(buffer, "ADDRESS A%d;\n", i);
263 }
264
265 for(i = 0; i < This->baseShader.limits.texcoord; i++) {
266 if (reg_maps->texcoord[i])
267 shader_addline(buffer,"TEMP T%u;\n", i);
268 }
269
270 /* Texture coordinate registers must be pre-loaded */
271 for (i = 0; i < This->baseShader.limits.texcoord; i++) {
272 if (reg_maps->texcoord[i])
273 shader_addline(buffer, "MOV T%u, fragment.texcoord[%u];\n", i, i);
274 }
275
276 for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
277 IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) This;
278 if(!reg_maps->bumpmat[i]) continue;
279
280 cur = ps->numbumpenvmatconsts;
281 ps->bumpenvmatconst[cur].const_num = -1;
282 ps->bumpenvmatconst[cur].texunit = i;
283 ps->luminanceconst[cur].const_num = -1;
284 ps->luminanceconst[cur].texunit = i;
285
286 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
287 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
288 * bump mapping.
289 */
290 if(max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
291 ps->bumpenvmatconst[cur].const_num = max_constantsF + extra_constants_needed;
292 shader_addline(buffer, "PARAM bumpenvmat%d = program.env[%d];\n",
293 i, ps->bumpenvmatconst[cur].const_num);
294 extra_constants_needed++;
295
296 if(reg_maps->luminanceparams && max_constantsF + extra_constants_needed < GL_LIMITS(pshader_constantsF)) {
297 ((IWineD3DPixelShaderImpl *)This)->luminanceconst[cur].const_num = max_constantsF + extra_constants_needed;
298 shader_addline(buffer, "PARAM luminance%d = program.env[%d];\n",
299 i, ps->luminanceconst[cur].const_num);
300 extra_constants_needed++;
301 } else if(reg_maps->luminanceparams) {
302 FIXME("No free constant to load the luminance parameters\n");
303 }
304 } else {
305 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
306 }
307
308 ps->numbumpenvmatconsts = cur + 1;
309 }
310
311 if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE] && pshader) {
312 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
313 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
314 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
315 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
316 * off again
317 */
318 if(max_constantsF + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF) && FALSE) {
319 /* The idea is that if srgb is enabled, then disabled, the constant loading code
320 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
321 * and comparison constants. If it disables it that way, the shader won't be recompiled
322 * and the code will stay in, so sRGB writing can be turned on again by setting the
323 * constants from the spec
324 */
325 ps_impl->srgb_mode_hardcoded = 0;
326 ps_impl->srgb_low_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed;
327 ps_impl->srgb_cmp_const = GL_LIMITS(pshader_constantsF) - extra_constants_needed - 1;
328 shader_addline(buffer, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl->srgb_low_const);
329 shader_addline(buffer, "PARAM srgb_comparison = program.env[%d];\n", ps_impl->srgb_cmp_const);
330 } else {
331 shader_addline(buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
332 srgb_mul_low, srgb_mul_low, srgb_mul_low);
333 shader_addline(buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
334 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
335 ps_impl->srgb_mode_hardcoded = 1;
336 }
337 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
338 * path if the comparison value is set to INF
339 */
340 shader_addline(buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
341 srgb_pow, srgb_pow, srgb_pow);
342 shader_addline(buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
343 srgb_mul_high, srgb_mul_high, srgb_mul_high);
344 shader_addline(buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
345 srgb_sub_high, srgb_sub_high, srgb_sub_high);
346 ps_impl->srgb_enabled = 1;
347 } else if(pshader) {
348 IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
349
350 /* Do not write any srgb fixup into the shader to save shader size and processing time.
351 * As a consequence, we can't toggle srgb write on without recompilation
352 */
353 ps_impl->srgb_enabled = 0;
354 ps_impl->srgb_mode_hardcoded = 1;
355 }
356
357 /* Load local constants using the program-local space,
358 * this avoids reloading them each time the shader is used
359 */
360 if(!This->baseShader.load_local_constsF) {
361 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
362 shader_addline(buffer, "PARAM C%u = program.local[%u];\n", lconst->idx,
363 lconst->idx);
364 }
365 }
366
367 /* we use the array-based constants array if the local constants are marked for loading,
368 * because then we use indirect addressing, or when the local constant list is empty,
369 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
370 * local constants do not declare the loaded constants as an array because ARB compilers usually
371 * do not optimize unused constants away
372 */
373 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
374 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
375 shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
376 max_constantsF, max_constantsF - 1);
377 } else {
378 for(i = 0; i < max_constantsF; i++) {
379 if(!shader_constant_is_local(This, i)) {
380 shader_addline(buffer, "PARAM C%d = program.env[%d];\n",i, i);
381 }
382 }
383 }
384 }
385
386 static const char * const shift_tab[] = {
387 "dummy", /* 0 (none) */
388 "coefmul.x", /* 1 (x2) */
389 "coefmul.y", /* 2 (x4) */
390 "coefmul.z", /* 3 (x8) */
391 "coefmul.w", /* 4 (x16) */
392 "dummy", /* 5 (x32) */
393 "dummy", /* 6 (x64) */
394 "dummy", /* 7 (x128) */
395 "dummy", /* 8 (d256) */
396 "dummy", /* 9 (d128) */
397 "dummy", /* 10 (d64) */
398 "dummy", /* 11 (d32) */
399 "coefdiv.w", /* 12 (d16) */
400 "coefdiv.z", /* 13 (d8) */
401 "coefdiv.y", /* 14 (d4) */
402 "coefdiv.x" /* 15 (d2) */
403 };
404
405 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG* arg, const DWORD param, char *write_mask) {
406 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
407 char *ptr = write_mask;
408 char vshader = shader_is_vshader_version(This->baseShader.hex_version);
409
410 if(vshader && shader_get_regtype(param) == WINED3DSPR_ADDR) {
411 *ptr++ = '.';
412 *ptr++ = 'x';
413 } else if ((param & WINED3DSP_WRITEMASK_ALL) != WINED3DSP_WRITEMASK_ALL) {
414 *ptr++ = '.';
415 if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
416 if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
417 if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
418 if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
419 }
420
421 *ptr = '\0';
422 }
423
424 static void shader_arb_get_swizzle(const DWORD param, BOOL fixup, char *swizzle_str) {
425 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
426 * but addressed as "rgba". To fix this we need to swap the register's x
427 * and z components. */
428 const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
429 char *ptr = swizzle_str;
430
431 /* swizzle bits fields: wwzzyyxx */
432 DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
433 DWORD swizzle_x = swizzle & 0x03;
434 DWORD swizzle_y = (swizzle >> 2) & 0x03;
435 DWORD swizzle_z = (swizzle >> 4) & 0x03;
436 DWORD swizzle_w = (swizzle >> 6) & 0x03;
437
438 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
439 * generate a swizzle string. Unless we need to our own swizzling. */
440 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) != swizzle || fixup) {
441 *ptr++ = '.';
442 if (swizzle_x == swizzle_y && swizzle_x == swizzle_z && swizzle_x == swizzle_w) {
443 *ptr++ = swizzle_chars[swizzle_x];
444 } else {
445 *ptr++ = swizzle_chars[swizzle_x];
446 *ptr++ = swizzle_chars[swizzle_y];
447 *ptr++ = swizzle_chars[swizzle_z];
448 *ptr++ = swizzle_chars[swizzle_w];
449 }
450 }
451
452 *ptr = '\0';
453 }
454
455 static void pshader_get_register_name(IWineD3DBaseShader* iface,
456 const DWORD param, char* regstr) {
457
458 DWORD reg = param & WINED3DSP_REGNUM_MASK;
459 DWORD regtype = shader_get_regtype(param);
460 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
461
462 switch (regtype) {
463 case WINED3DSPR_TEMP:
464 sprintf(regstr, "R%u", reg);
465 break;
466 case WINED3DSPR_INPUT:
467 if (reg==0) {
468 strcpy(regstr, "fragment.color.primary");
469 } else {
470 strcpy(regstr, "fragment.color.secondary");
471 }
472 break;
473 case WINED3DSPR_CONST:
474 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
475 sprintf(regstr, "C[%u]", reg);
476 } else {
477 sprintf(regstr, "C%u", reg);
478 }
479 break;
480 case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
481 sprintf(regstr,"T%u", reg);
482 break;
483 case WINED3DSPR_COLOROUT:
484 if (reg == 0)
485 sprintf(regstr, "TMP_COLOR");
486 else {
487 /* TODO: See GL_ARB_draw_buffers */
488 FIXME("Unsupported write to render target %u\n", reg);
489 sprintf(regstr, "unsupported_register");
490 }
491 break;
492 case WINED3DSPR_DEPTHOUT:
493 sprintf(regstr, "result.depth");
494 break;
495 case WINED3DSPR_ATTROUT:
496 sprintf(regstr, "oD[%u]", reg);
497 break;
498 case WINED3DSPR_TEXCRDOUT:
499 sprintf(regstr, "oT[%u]", reg);
500 break;
501 default:
502 FIXME("Unhandled register name Type(%d)\n", regtype);
503 sprintf(regstr, "unrecognized_register");
504 break;
505 }
506 }
507
508 /* TODO: merge with pixel shader */
509 static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param, BOOL is_input, char *hwLine) {
510
511 IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
512
513 /* oPos, oFog and oPts in D3D */
514 static const char * const hwrastout_reg_names[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
515
516 DWORD reg = param & WINED3DSP_REGNUM_MASK;
517 DWORD regtype = shader_get_regtype(param);
518 char tmpReg[255];
519 BOOL is_color = FALSE;
520
521 if ((param & WINED3DSP_SRCMOD_MASK) == WINED3DSPSM_NEG) {
522 strcat(hwLine, " -");
523 } else {
524 strcat(hwLine, " ");
525 }
526
527 switch (regtype) {
528 case WINED3DSPR_TEMP:
529 sprintf(tmpReg, "R%u", reg);
530 strcat(hwLine, tmpReg);
531 break;
532 case WINED3DSPR_INPUT:
533
534 if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
535 is_color = TRUE;
536
537 sprintf(tmpReg, "vertex.attrib[%u]", reg);
538 strcat(hwLine, tmpReg);
539 break;
540 case WINED3DSPR_CONST:
541 if(param & WINED3DSHADER_ADDRMODE_RELATIVE) {
542 if(reg >= This->rel_offset) {
543 sprintf(tmpReg, "C[A0.x + %u]", reg - This->rel_offset);
544 } else {
545 sprintf(tmpReg, "C[A0.x - %u]", -reg + This->rel_offset);
546 }
547 } else {
548 if(This->baseShader.load_local_constsF || list_empty(&This->baseShader.constantsF)) {
549 sprintf(tmpReg, "C[%u]", reg);
550 } else {
551 sprintf(tmpReg, "C%u", reg);
552 }
553 }
554 strcat(hwLine, tmpReg);
555 break;
556 case WINED3DSPR_ADDR: /*case D3DSPR_TEXTURE:*/
557 sprintf(tmpReg, "A%u", reg);
558 strcat(hwLine, tmpReg);
559 break;
560 case WINED3DSPR_RASTOUT:
561 sprintf(tmpReg, "%s", hwrastout_reg_names[reg]);
562 strcat(hwLine, tmpReg);
563 break;
564 case WINED3DSPR_ATTROUT:
565 if (reg==0) {
566 strcat(hwLine, "result.color.primary");
567 } else {
568 strcat(hwLine, "result.color.secondary");
569 }
570 break;
571 case WINED3DSPR_TEXCRDOUT:
572 sprintf(tmpReg, "result.texcoord[%u]", reg);
573 strcat(hwLine, tmpReg);
574 break;
575 default:
576 FIXME("Unknown reg type %d %d\n", regtype, reg);
577 strcat(hwLine, "unrecognized_register");
578 break;
579 }
580
581 if (!is_input) {
582 char write_mask[6];
583 shader_arb_get_write_mask(arg, param, write_mask);
584 strcat(hwLine, write_mask);
585 } else {
586 char swizzle[6];
587 shader_arb_get_swizzle(param, is_color, swizzle);
588 strcat(hwLine, swizzle);
589 }
590 }
591
592 static void shader_hw_sample(SHADER_OPCODE_ARG* arg, DWORD sampler_idx, const char *dst_str, const char *coord_reg, BOOL projected, BOOL bias) {
593 SHADER_BUFFER* buffer = arg->buffer;
594 DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
595 const char *tex_type;
596
597 switch(sampler_type) {
598 case WINED3DSTT_1D:
599 tex_type = "1D";
600 break;
601
602 case WINED3DSTT_2D:
603 {
604 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) arg->shader;
605 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
606 if(device->stateBlock->textures[sampler_idx] &&
607 IWineD3DBaseTexture_GetTextureDimensions(device->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
608 tex_type = "RECT";
609 } else {
610 tex_type = "2D";
611 }
612 break;
613 }
614
615 case WINED3DSTT_VOLUME:
616 tex_type = "3D";
617 break;
618
619 case WINED3DSTT_CUBE:
620 tex_type = "CUBE";
621 break;
622
623 default:
624 ERR("Unexpected texture type %d\n", sampler_type);
625 tex_type = "";
626 }
627
628 if (bias) {
629 /* Shouldn't be possible, but let's check for it */
630 if(projected) FIXME("Biased and Projected texture sampling\n");
631 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
632 shader_addline(buffer, "TXB %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
633 } else if (projected) {
634 shader_addline(buffer, "TXP %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
635 } else {
636 shader_addline(buffer, "TEX %s, %s, texture[%u], %s;\n", dst_str, coord_reg, sampler_idx, tex_type);
637 }
638 }
639
640 static void gen_color_correction(SHADER_BUFFER *buffer, const char *reg, const char *writemask,
641 const char *one, const char *two, WINED3DFORMAT fmt,
642 WineD3D_GL_Info *gl_info) {
643 switch(fmt) {
644 case WINED3DFMT_V8U8:
645 case WINED3DFMT_V16U16:
646 if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
647 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
648 if(0) {
649 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
650 * disabled until an application that needs it is found because it causes unneeded
651 * shader recompilation in some game
652 */
653 if(strlen(writemask) >= 4) {
654 shader_addline(buffer, "MOV %s.%c, %s;\n", reg, writemask[3], one);
655 }
656 }
657 } else {
658 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
659 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
660 * all registers, do so, this saves an instruction.
661 */
662 if(strlen(writemask) >= 5) {
663 shader_addline(buffer, "MAD %s, %s, %s, -%s;\n", reg, reg, two, one);
664 } else if(strlen(writemask) >= 3) {
665 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
666 reg, writemask[1],
667 reg, writemask[1],
668 two, one);
669 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
670 reg, writemask[2],
671 reg, writemask[2],
672 two, one);
673 } else if(strlen(writemask) == 2) {
674 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n", reg, writemask[1],
675 reg, writemask[1], two, one);
676 }
677 }
678 break;
679
680 case WINED3DFMT_X8L8V8U8:
681 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
682 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
683 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
684 */
685 if(strlen(writemask) >= 3) {
686 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
687 reg, writemask[1],
688 reg, writemask[1],
689 two, one);
690 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
691 reg, writemask[2],
692 reg, writemask[2],
693 two, one);
694 } else if(strlen(writemask) == 2) {
695 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
696 reg, writemask[1],
697 reg, writemask[1],
698 two, one);
699 }
700 }
701 break;
702
703 case WINED3DFMT_L6V5U5:
704 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
705 if(strlen(writemask) >= 4) {
706 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
707 shader_addline(buffer, "MOV TMP.g, %s.%c;\n",
708 reg, writemask[2]);
709 shader_addline(buffer, "MAD %s.%c%c, %s.%c%c, %s, -%s;\n",
710 reg, writemask[1], writemask[1],
711 reg, writemask[1], writemask[3],
712 two, one);
713 shader_addline(buffer, "MOV %s.%c, TMP.g;\n", reg,
714 writemask[3]);
715 } else if(strlen(writemask) == 3) {
716 /* This is bad: We have VL, but we need VU */
717 FIXME("2 components sampled from a converted L6V5U5 texture\n");
718 } else {
719 shader_addline(buffer, "MAD %s.%c, %s.%c, %s, -%s;\n",
720 reg, writemask[1],
721 reg, writemask[1],
722 two, one);
723 }
724 }
725 break;
726
727 case WINED3DFMT_Q8W8V8U8:
728 if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
729 /* Correct the sign in all channels */
730 switch(strlen(writemask)) {
731 case 4:
732 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
733 reg, writemask[3],
734 reg, writemask[3]);
735 /* drop through */
736 case 3:
737 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
738 reg, writemask[2],
739 reg, writemask[2]);
740 /* drop through */
741 case 2:
742 shader_addline(buffer, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
743 reg, writemask[1],
744 reg, writemask[1]);
745 break;
746
747 /* Should not occur, since it's at minimum '.' and a letter */
748 case 1:
749 ERR("Unexpected writemask: \"%s\"\n", writemask);
750 break;
751
752 case 5:
753 default:
754 shader_addline(buffer, "MAD %s, %s, coefmul.x, -one;\n", reg, reg);
755 }
756 }
757 break;
758
759 case WINED3DFMT_ATI2N:
760 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
761 * which means the first one is replicated across .rgb, and the 2nd one is in
762 * .a. We need the 2nd in .g
763 *
764 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
765 * are swapped compared to d3d. So swap red and green.
766 */
767 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC)) {
768 shader_addline(buffer, "SWZ %s, %s, %c, %c, 1, 0;\n",
769 reg, reg, writemask[2], writemask[1]);
770 } else {
771 if(strlen(writemask) == 5) {
772 shader_addline(buffer, "MOV %s.%c, %s.%c;\n",
773 reg, writemask[2], reg, writemask[4]);
774 } else if(strlen(writemask) == 2) {
775 /* Nothing to do */
776 } else {
777 /* This is bad: We have VL, but we need VU */
778 FIXME("2 or 3 components sampled from a converted ATI2N texture\n");
779 }
780 }
781 break;
782
783 /* stupid compiler */
784 default:
785 break;
786 }
787 }
788
789 static void shader_arb_color_correction(SHADER_OPCODE_ARG* arg) {
790 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
791 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
792 WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
793 WINED3DFORMAT fmt;
794 WINED3DFORMAT conversion_group;
795 IWineD3DBaseTextureImpl *texture;
796 UINT i;
797 BOOL recorded = FALSE;
798 DWORD sampler_idx;
799 DWORD hex_version = shader->baseShader.hex_version;
800 char reg[256];
801 char writemask[6];
802
803 switch(arg->opcode->opcode) {
804 case WINED3DSIO_TEX:
805 if (hex_version < WINED3DPS_VERSION(2,0)) {
806 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
807 } else {
808 sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
809 }
810 break;
811
812 case WINED3DSIO_TEXLDL:
813 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
814 return;
815
816 case WINED3DSIO_TEXDP3TEX:
817 case WINED3DSIO_TEXM3x3TEX:
818 case WINED3DSIO_TEXM3x3SPEC:
819 case WINED3DSIO_TEXM3x3VSPEC:
820 case WINED3DSIO_TEXBEM:
821 case WINED3DSIO_TEXREG2AR:
822 case WINED3DSIO_TEXREG2GB:
823 case WINED3DSIO_TEXREG2RGB:
824 sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
825 break;
826
827 default:
828 /* Not a texture sampling instruction, nothing to do */
829 return;
830 };
831
832 texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
833 if(texture) {
834 fmt = texture->resource.format;
835 conversion_group = texture->baseTexture.shader_conversion_group;
836 } else {
837 fmt = WINED3DFMT_UNKNOWN;
838 conversion_group = WINED3DFMT_UNKNOWN;
839 }
840
841 /* before doing anything, record the sampler with the format in the format conversion list,
842 * but check if it's not there already
843 */
844 for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
845 if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
846 recorded = TRUE;
847 }
848 }
849 if(!recorded) {
850 shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
851 shader->baseShader.num_sampled_samplers++;
852 shader->baseShader.sampled_format[sampler_idx] = conversion_group;
853 }
854
855 pshader_get_register_name(arg->shader, arg->dst, reg);
856 shader_arb_get_write_mask(arg, arg->dst, writemask);
857 if(strlen(writemask) == 0) strcpy(writemask, ".xyzw");
858
859 gen_color_correction(arg->buffer, reg, writemask, "one", "coefmul.x", fmt, gl_info);
860
861 }
862
863
864 static void pshader_gen_input_modifier_line (
865 IWineD3DBaseShader *iface,
866 SHADER_BUFFER* buffer,
867 const DWORD instr,
868 int tmpreg,
869 char *outregstr) {
870
871 /* Generate a line that does the input modifier computation and return the input register to use */
872 char regstr[256];
873 char swzstr[20];
874 int insert_line;
875
876 /* Assume a new line will be added */
877 insert_line = 1;
878
879 /* Get register name */
880 pshader_get_register_name(iface, instr, regstr);
881 shader_arb_get_swizzle(instr, FALSE, swzstr);
882
883 switch (instr & WINED3DSP_SRCMOD_MASK) {
884 case WINED3DSPSM_NONE:
885 sprintf(outregstr, "%s%s", regstr, swzstr);
886 insert_line = 0;
887 break;
888 case WINED3DSPSM_NEG:
889 sprintf(outregstr, "-%s%s", regstr, swzstr);
890 insert_line = 0;
891 break;
892 case WINED3DSPSM_BIAS:
893 shader_addline(buffer, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg, regstr);
894 break;
895 case WINED3DSPSM_BIASNEG:
896 shader_addline(buffer, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg, regstr);
897 break;
898 case WINED3DSPSM_SIGN:
899 shader_addline(buffer, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg, regstr);
900 break;
901 case WINED3DSPSM_SIGNNEG:
902 shader_addline(buffer, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg, regstr);
903 break;
904 case WINED3DSPSM_COMP:
905 shader_addline(buffer, "SUB T%c, one.x, %s;\n", 'A' + tmpreg, regstr);
906 break;
907 case WINED3DSPSM_X2:
908 shader_addline(buffer, "ADD T%c, %s, %s;\n", 'A' + tmpreg, regstr, regstr);
909 break;
910 case WINED3DSPSM_X2NEG:
911 shader_addline(buffer, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg, regstr, regstr);
912 break;
913 case WINED3DSPSM_DZ:
914 shader_addline(buffer, "RCP T%c, %s.z;\n", 'A' + tmpreg, regstr);
915 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
916 break;
917 case WINED3DSPSM_DW:
918 shader_addline(buffer, "RCP T%c, %s.w;\n", 'A' + tmpreg, regstr);
919 shader_addline(buffer, "MUL T%c, %s, T%c;\n", 'A' + tmpreg, regstr, 'A' + tmpreg);
920 break;
921 default:
922 sprintf(outregstr, "%s%s", regstr, swzstr);
923 insert_line = 0;
924 }
925
926 /* Return modified or original register, with swizzle */
927 if (insert_line)
928 sprintf(outregstr, "T%c%s", 'A' + tmpreg, swzstr);
929 }
930
931 static inline void pshader_gen_output_modifier_line(
932 SHADER_BUFFER* buffer,
933 int saturate,
934 char *write_mask,
935 int shift,
936 char *regstr) {
937
938 /* Generate a line that does the output modifier computation */
939 shader_addline(buffer, "MUL%s %s%s, %s, %s;\n", saturate ? "_SAT" : "",
940 regstr, write_mask, regstr, shift_tab[shift]);
941 }
942
943 void pshader_hw_bem(SHADER_OPCODE_ARG* arg) {
944 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
945
946 SHADER_BUFFER* buffer = arg->buffer;
947 char dst_name[50];
948 char src_name[2][50];
949 char dst_wmask[20];
950 DWORD sampler_code = arg->dst & WINED3DSP_REGNUM_MASK;
951 BOOL has_bumpmat = FALSE;
952 int i;
953
954 for(i = 0; i < This->numbumpenvmatconsts; i++) {
955 if(This->bumpenvmatconst[i].const_num != -1 && This->bumpenvmatconst[i].texunit == sampler_code) {
956 has_bumpmat = TRUE;
957 break;
958 }
959 }
960
961 pshader_get_register_name(arg->shader, arg->dst, dst_name);
962 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
963 strcat(dst_name, dst_wmask);
964
965 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
966 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
967
968 if(has_bumpmat) {
969 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
970 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code);
971 shader_addline(buffer, "DP3 TMP.r, TMP2, %s;\n", src_name[1]);
972 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code);
973 shader_addline(buffer, "DP3 TMP.g, TMP2, %s;\n", src_name[1]);
974
975 shader_addline(buffer, "ADD %s, %s, TMP;\n", dst_name, src_name[0]);
976 } else {
977 shader_addline(buffer, "MOV %s, %s;\n", dst_name, src_name[0]);
978 }
979 }
980
981 void pshader_hw_cnd(SHADER_OPCODE_ARG* arg) {
982
983 IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
984 SHADER_BUFFER* buffer = arg->buffer;
985 char dst_wmask[20];
986 char dst_name[50];
987 char src_name[3][50];
988 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
989 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
990
991 /* FIXME: support output modifiers */
992
993 /* Handle output register */
994 pshader_get_register_name(arg->shader, arg->dst, dst_name);
995 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
996
997 /* Generate input register names (with modifiers) */
998 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
999 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1000 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1001
1002 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1003 if (shader->baseShader.hex_version <= WINED3DPS_VERSION(1, 3) &&
1004 arg->opcode_token & WINED3DSI_COISSUE) {
1005 shader_addline(buffer, "MOV%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1]);
1006 } else {
1007 shader_addline(buffer, "ADD TMP, -%s, coefdiv.x;\n", src_name[0]);
1008 shader_addline(buffer, "CMP%s %s%s, TMP, %s, %s;\n",
1009 sat ? "_SAT" : "", dst_name, dst_wmask, src_name[1], src_name[2]);
1010 }
1011 if (shift != 0)
1012 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1013 }
1014
1015 void pshader_hw_cmp(SHADER_OPCODE_ARG* arg) {
1016
1017 SHADER_BUFFER* buffer = arg->buffer;
1018 char dst_wmask[20];
1019 char dst_name[50];
1020 char src_name[3][50];
1021 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1022 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1023
1024 /* FIXME: support output modifiers */
1025
1026 /* Handle output register */
1027 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1028 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1029
1030 /* Generate input register names (with modifiers) */
1031 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1032 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1033 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1034
1035 shader_addline(buffer, "CMP%s %s%s, %s, %s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1036 src_name[0], src_name[2], src_name[1]);
1037
1038 if (shift != 0)
1039 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1040 }
1041
1042 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1043 * dst = dot2(src0, src1) + src2 */
1044 void pshader_hw_dp2add(SHADER_OPCODE_ARG* arg) {
1045 SHADER_BUFFER* buffer = arg->buffer;
1046 char dst_wmask[20];
1047 char dst_name[50];
1048 char src_name[3][50];
1049 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1050 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1051
1052 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1053 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1054
1055 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name[0]);
1056 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[1], 1, src_name[1]);
1057 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[2], 2, src_name[2]);
1058
1059 /* Emulate a DP2 with a DP3 and 0.0 */
1060 shader_addline(buffer, "MOV TMP, %s;\n", src_name[0]);
1061 shader_addline(buffer, "MOV TMP.z, 0.0;\n");
1062 shader_addline(buffer, "DP3 TMP2, TMP, %s;\n", src_name[1]);
1063 shader_addline(buffer, "ADD%s %s%s, TMP2, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask, src_name[2]);
1064
1065 if (shift != 0)
1066 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1067 }
1068
1069 /* Map the opcode 1-to-1 to the GL code */
1070 void pshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1071
1072 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1073 SHADER_BUFFER* buffer = arg->buffer;
1074 DWORD dst = arg->dst;
1075 DWORD* src = arg->src;
1076
1077 unsigned int i;
1078 char tmpLine[256];
1079
1080 /* Output token related */
1081 char output_rname[256];
1082 char output_wmask[20];
1083 BOOL saturate = FALSE;
1084 BOOL centroid = FALSE;
1085 BOOL partialprecision = FALSE;
1086 DWORD shift;
1087
1088 strcpy(tmpLine, curOpcode->glname);
1089
1090 /* Process modifiers */
1091 if (0 != (dst & WINED3DSP_DSTMOD_MASK)) {
1092 DWORD mask = dst & WINED3DSP_DSTMOD_MASK;
1093
1094 saturate = mask & WINED3DSPDM_SATURATE;
1095 centroid = mask & WINED3DSPDM_MSAMPCENTROID;
1096 partialprecision = mask & WINED3DSPDM_PARTIALPRECISION;
1097 mask &= ~(WINED3DSPDM_MSAMPCENTROID | WINED3DSPDM_PARTIALPRECISION | WINED3DSPDM_SATURATE);
1098 if (mask)
1099 FIXME("Unrecognized modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1100
1101 if (centroid)
1102 FIXME("Unhandled modifier(%#x)\n", mask >> WINED3DSP_DSTMOD_SHIFT);
1103 }
1104 shift = (dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1105
1106 /* Generate input and output registers */
1107 if (curOpcode->num_params > 0) {
1108 char operands[4][100];
1109
1110 /* Generate input register names (with modifiers) */
1111 for (i = 1; i < curOpcode->num_params; ++i)
1112 pshader_gen_input_modifier_line(arg->shader, buffer, src[i-1], i-1, operands[i]);
1113
1114 /* Handle output register */
1115 pshader_get_register_name(arg->shader, dst, output_rname);
1116 strcpy(operands[0], output_rname);
1117 shader_arb_get_write_mask(arg, dst, output_wmask);
1118 strcat(operands[0], output_wmask);
1119
1120 if (saturate && (shift == 0))
1121 strcat(tmpLine, "_SAT");
1122 strcat(tmpLine, " ");
1123 strcat(tmpLine, operands[0]);
1124 for (i = 1; i < curOpcode->num_params; i++) {
1125 strcat(tmpLine, ", ");
1126 strcat(tmpLine, operands[i]);
1127 }
1128 strcat(tmpLine,";\n");
1129 shader_addline(buffer, tmpLine);
1130
1131 /* A shift requires another line. */
1132 if (shift != 0)
1133 pshader_gen_output_modifier_line(buffer, saturate, output_wmask, shift, output_rname);
1134 }
1135 }
1136
1137 void pshader_hw_texkill(SHADER_OPCODE_ARG* arg) {
1138 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1139 DWORD hex_version = This->baseShader.hex_version;
1140 SHADER_BUFFER* buffer = arg->buffer;
1141 char reg_dest[40];
1142
1143 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1144 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1145 */
1146 pshader_get_register_name(arg->shader, arg->dst, reg_dest);
1147
1148 if(hex_version >= WINED3DPS_VERSION(2,0)) {
1149 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1150 shader_addline(buffer, "KIL %s;\n", reg_dest);
1151 } else {
1152 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1153 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1154 */
1155 shader_addline(buffer, "MOV TMP, %s;\n", reg_dest);
1156 shader_addline(buffer, "MOV TMP.w, one.w;\n");
1157 shader_addline(buffer, "KIL TMP;\n");
1158 }
1159 }
1160
1161 void pshader_hw_tex(SHADER_OPCODE_ARG* arg) {
1162 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1163 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1164
1165 DWORD dst = arg->dst;
1166 DWORD* src = arg->src;
1167 SHADER_BUFFER* buffer = arg->buffer;
1168 DWORD hex_version = This->baseShader.hex_version;
1169 BOOL projected = FALSE, bias = FALSE;
1170
1171 char reg_dest[40];
1172 char reg_coord[40];
1173 DWORD reg_dest_code;
1174 DWORD reg_sampler_code;
1175
1176 /* All versions have a destination register */
1177 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1178 pshader_get_register_name(arg->shader, dst, reg_dest);
1179
1180 /* 1.0-1.3: Use destination register as coordinate source.
1181 1.4+: Use provided coordinate source register. */
1182 if (hex_version < WINED3DPS_VERSION(1,4))
1183 strcpy(reg_coord, reg_dest);
1184 else
1185 pshader_gen_input_modifier_line(arg->shader, buffer, src[0], 0, reg_coord);
1186
1187 /* 1.0-1.4: Use destination register number as texture code.
1188 2.0+: Use provided sampler number as texure code. */
1189 if (hex_version < WINED3DPS_VERSION(2,0))
1190 reg_sampler_code = reg_dest_code;
1191 else
1192 reg_sampler_code = src[1] & WINED3DSP_REGNUM_MASK;
1193
1194 /* projection flag:
1195 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1196 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1197 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1198 */
1199 if(hex_version < WINED3DPS_VERSION(1,4)) {
1200 DWORD flags = 0;
1201 if(reg_sampler_code < MAX_TEXTURES) {
1202 flags = deviceImpl->stateBlock->textureState[reg_sampler_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1203 }
1204 if (flags & WINED3DTTFF_PROJECTED) {
1205 projected = TRUE;
1206 }
1207 } else if(hex_version < WINED3DPS_VERSION(2,0)) {
1208 DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
1209 if (src_mod == WINED3DSPSM_DZ) {
1210 projected = TRUE;
1211 } else if(src_mod == WINED3DSPSM_DW) {
1212 projected = TRUE;
1213 }
1214 } else {
1215 if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
1216 projected = TRUE;
1217 }
1218 if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
1219 bias = TRUE;
1220 }
1221 }
1222 shader_hw_sample(arg, reg_sampler_code, reg_dest, reg_coord, projected, bias);
1223 }
1224
1225 void pshader_hw_texcoord(SHADER_OPCODE_ARG* arg) {
1226
1227 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1228 DWORD dst = arg->dst;
1229 SHADER_BUFFER* buffer = arg->buffer;
1230 DWORD hex_version = This->baseShader.hex_version;
1231
1232 char tmp[20];
1233 shader_arb_get_write_mask(arg, dst, tmp);
1234 if (hex_version != WINED3DPS_VERSION(1,4)) {
1235 DWORD reg = dst & WINED3DSP_REGNUM_MASK;
1236 shader_addline(buffer, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg, tmp, reg);
1237 } else {
1238 DWORD reg1 = dst & WINED3DSP_REGNUM_MASK;
1239 char reg_src[40];
1240
1241 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, reg_src);
1242 shader_addline(buffer, "MOV R%u%s, %s;\n", reg1, tmp, reg_src);
1243 }
1244 }
1245
1246 void pshader_hw_texreg2ar(SHADER_OPCODE_ARG* arg) {
1247
1248 SHADER_BUFFER* buffer = arg->buffer;
1249 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1250 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1251 DWORD flags;
1252
1253 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1254 char dst_str[8];
1255 char src_str[50];
1256
1257 sprintf(dst_str, "T%u", reg1);
1258 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1259 shader_addline(buffer, "MOV TMP.r, %s.a;\n", src_str);
1260 shader_addline(buffer, "MOV TMP.g, %s.r;\n", src_str);
1261 flags = reg1 < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg1][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1262 shader_hw_sample(arg, reg1, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1263 }
1264
1265 void pshader_hw_texreg2gb(SHADER_OPCODE_ARG* arg) {
1266
1267 SHADER_BUFFER* buffer = arg->buffer;
1268
1269 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1270 char dst_str[8];
1271 char src_str[50];
1272
1273 sprintf(dst_str, "T%u", reg1);
1274 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1275 shader_addline(buffer, "MOV TMP.r, %s.g;\n", src_str);
1276 shader_addline(buffer, "MOV TMP.g, %s.b;\n", src_str);
1277 shader_hw_sample(arg, reg1, dst_str, "TMP", FALSE, FALSE);
1278 }
1279
1280 void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG* arg) {
1281
1282 SHADER_BUFFER* buffer = arg->buffer;
1283 DWORD reg1 = arg->dst & WINED3DSP_REGNUM_MASK;
1284 char dst_str[8];
1285 char src_str[50];
1286
1287 sprintf(dst_str, "T%u", reg1);
1288 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_str);
1289 shader_hw_sample(arg, reg1, dst_str, src_str, FALSE, FALSE);
1290 }
1291
1292 void pshader_hw_texbem(SHADER_OPCODE_ARG* arg) {
1293 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1294 BOOL has_bumpmat = FALSE;
1295 BOOL has_luminance = FALSE;
1296 int i;
1297
1298 DWORD dst = arg->dst;
1299 DWORD src = arg->src[0] & WINED3DSP_REGNUM_MASK;
1300 SHADER_BUFFER* buffer = arg->buffer;
1301
1302 char reg_coord[40];
1303 DWORD reg_dest_code;
1304
1305 /* All versions have a destination register */
1306 reg_dest_code = dst & WINED3DSP_REGNUM_MASK;
1307 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1308 pshader_get_register_name(arg->shader, dst, reg_coord);
1309
1310 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1311 if(This->bumpenvmatconst[i].const_num != -1 && reg_dest_code == This->bumpenvmatconst[i].texunit) {
1312 has_bumpmat = TRUE;
1313 break;
1314 }
1315 }
1316 for(i = 0; i < This->numbumpenvmatconsts; i++) {
1317 if(This->luminanceconst[i].const_num != -1 && reg_dest_code == This->luminanceconst[i].texunit) {
1318 has_luminance = TRUE;
1319 break;
1320 }
1321 }
1322
1323 if(has_bumpmat) {
1324 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1325
1326 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code);
1327 shader_addline(buffer, "DP3 TMP.r, TMP2, T%u;\n", src);
1328 shader_addline(buffer, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code);
1329 shader_addline(buffer, "DP3 TMP.g, TMP2, T%u;\n", src);
1330
1331 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1332 * so we can't let the GL handle this.
1333 */
1334 if (((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS]
1335 & WINED3DTTFF_PROJECTED) {
1336 shader_addline(buffer, "RCP TMP2.a, %s.a;\n", reg_coord);
1337 shader_addline(buffer, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord);
1338 shader_addline(buffer, "ADD TMP.rg, TMP, TMP2;\n");
1339 } else {
1340 shader_addline(buffer, "ADD TMP.rg, TMP, %s;\n", reg_coord);
1341 }
1342
1343 shader_hw_sample(arg, reg_dest_code, reg_coord, "TMP", FALSE, FALSE);
1344
1345 if(arg->opcode->opcode == WINED3DSIO_TEXBEML && has_luminance) {
1346 shader_addline(buffer, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1347 src, reg_dest_code, reg_dest_code);
1348 shader_addline(buffer, "MUL %s, %s, TMP;\n", reg_coord, reg_coord);
1349 }
1350
1351 } else {
1352 DWORD tf;
1353 if(reg_dest_code < MAX_TEXTURES) {
1354 tf = ((IWineD3DDeviceImpl*) This->baseShader.device)->stateBlock->textureState[reg_dest_code][WINED3DTSS_TEXTURETRANSFORMFLAGS];
1355 } else {
1356 tf = 0;
1357 }
1358 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1359 shader_hw_sample(arg, reg_dest_code, reg_coord, reg_coord, tf & WINED3DTTFF_PROJECTED, FALSE);
1360 }
1361 }
1362
1363 void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG* arg) {
1364
1365 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1366 SHADER_BUFFER* buffer = arg->buffer;
1367 char src0_name[50];
1368
1369 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1370 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", reg, src0_name);
1371 }
1372
1373 void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG* arg) {
1374
1375 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1376 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1377 DWORD flags;
1378 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1379 SHADER_BUFFER* buffer = arg->buffer;
1380 char dst_str[8];
1381 char src0_name[50];
1382
1383 sprintf(dst_str, "T%u", reg);
1384 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1385 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", reg, src0_name);
1386 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1387 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1388 }
1389
1390 void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG* arg) {
1391
1392 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1393 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1394 SHADER_BUFFER* buffer = arg->buffer;
1395 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1396 char src0_name[50];
1397
1398 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1399 shader_addline(buffer, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state->current_row, reg, src0_name);
1400 current_state->texcoord_w[current_state->current_row++] = reg;
1401 }
1402
1403 void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG* arg) {
1404
1405 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1406 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1407 DWORD flags;
1408 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1409 SHADER_BUFFER* buffer = arg->buffer;
1410 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1411 char dst_str[8];
1412 char src0_name[50];
1413
1414 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1415 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1416
1417 /* Sample the texture using the calculated coordinates */
1418 sprintf(dst_str, "T%u", reg);
1419 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1420 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1421 current_state->current_row = 0;
1422 }
1423
1424 void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
1425
1426 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1427 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1428 DWORD flags;
1429 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1430 SHADER_BUFFER* buffer = arg->buffer;
1431 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1432 char dst_str[8];
1433 char src0_name[50];
1434
1435 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1436 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1437
1438 /* Construct the eye-ray vector from w coordinates */
1439 shader_addline(buffer, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state->texcoord_w[0]);
1440 shader_addline(buffer, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state->texcoord_w[1]);
1441 shader_addline(buffer, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg);
1442
1443 /* Calculate reflection vector
1444 */
1445 shader_addline(buffer, "DP3 TMP.w, TMP, TMP2;\n");
1446 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1447 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1448 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1449 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1450 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1451 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1452
1453 /* Sample the texture using the calculated coordinates */
1454 sprintf(dst_str, "T%u", reg);
1455 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1456 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1457 current_state->current_row = 0;
1458 }
1459
1460 void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG* arg) {
1461
1462 IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
1463 IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
1464 DWORD flags;
1465 DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
1466 DWORD reg3 = arg->src[1] & WINED3DSP_REGNUM_MASK;
1467 SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
1468 SHADER_BUFFER* buffer = arg->buffer;
1469 char dst_str[8];
1470 char src0_name[50];
1471
1472 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0_name);
1473 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", reg, src0_name);
1474
1475 /* Calculate reflection vector.
1476 *
1477 * dot(N, E)
1478 * TMP.xyz = 2 * --------- * N - E
1479 * dot(N, N)
1480 *
1481 * Which normalizes the normal vector
1482 */
1483 shader_addline(buffer, "DP3 TMP.w, TMP, C[%u];\n", reg3);
1484 shader_addline(buffer, "DP3 TMP2.w, TMP, TMP;\n");
1485 shader_addline(buffer, "RCP TMP2.w, TMP2.w;\n");
1486 shader_addline(buffer, "MUL TMP.w, TMP.w, TMP2.w;\n");
1487 shader_addline(buffer, "MUL TMP, TMP.w, TMP;\n");
1488 shader_addline(buffer, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3);
1489
1490 /* Sample the texture using the calculated coordinates */
1491 sprintf(dst_str, "T%u", reg);
1492 flags = reg < MAX_TEXTURES ? deviceImpl->stateBlock->textureState[reg][WINED3DTSS_TEXTURETRANSFORMFLAGS] : 0;
1493 shader_hw_sample(arg, reg, dst_str, "TMP", flags & WINED3DTTFF_PROJECTED, FALSE);
1494 current_state->current_row = 0;
1495 }
1496
1497 void pshader_hw_texdepth(SHADER_OPCODE_ARG* arg) {
1498 SHADER_BUFFER* buffer = arg->buffer;
1499 char dst_name[50];
1500
1501 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1502 * which is essentially an input, is the destination register because it is the first
1503 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1504 * here
1505 */
1506 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1507
1508 /* According to the msdn, the source register(must be r5) is unusable after
1509 * the texdepth instruction, so we're free to modify it
1510 */
1511 shader_addline(buffer, "MIN %s.g, %s.g, one.g;\n", dst_name, dst_name);
1512
1513 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1514 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1515 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1516 */
1517 shader_addline(buffer, "RCP %s.g, %s.g;\n", dst_name, dst_name);
1518 shader_addline(buffer, "MUL TMP.x, %s.r, %s.g;\n", dst_name, dst_name);
1519 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1520 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1521 }
1522
1523 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1524 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1525 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1526 void pshader_hw_texdp3tex(SHADER_OPCODE_ARG* arg) {
1527 SHADER_BUFFER* buffer = arg->buffer;
1528 DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
1529 char src0[50];
1530 char dst_str[8];
1531
1532 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1533 shader_addline(buffer, "MOV TMP, 0.0;\n");
1534 shader_addline(buffer, "DP3 TMP.x, T%u, %s;\n", sampler_idx, src0);
1535
1536 sprintf(dst_str, "T%u", sampler_idx);
1537 shader_hw_sample(arg, sampler_idx, dst_str, "TMP", FALSE /* Only one coord, can't be projected */, FALSE);
1538 }
1539
1540 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1541 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1542 void pshader_hw_texdp3(SHADER_OPCODE_ARG* arg) {
1543 char src0[50];
1544 char dst_str[50];
1545 char dst_mask[6];
1546 DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
1547 SHADER_BUFFER* buffer = arg->buffer;
1548
1549 /* Handle output register */
1550 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1551 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1552
1553 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1554 shader_addline(buffer, "DP3 %s%s, T%u, %s;\n", dst_str, dst_mask, dstreg, src0);
1555
1556 /* TODO: Handle output modifiers */
1557 }
1558
1559 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1560 * Perform the 3rd row of a 3x3 matrix multiply */
1561 void pshader_hw_texm3x3(SHADER_OPCODE_ARG* arg) {
1562 SHADER_BUFFER* buffer = arg->buffer;
1563 char dst_str[50];
1564 char dst_mask[6];
1565 char src0[50];
1566 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1567
1568 pshader_get_register_name(arg->shader, arg->dst, dst_str);
1569 shader_arb_get_write_mask(arg, arg->dst, dst_mask);
1570
1571 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1572 shader_addline(buffer, "DP3 TMP.z, T%u, %s;\n", dst_reg, src0);
1573 shader_addline(buffer, "MOV %s%s, TMP;\n", dst_str, dst_mask);
1574
1575 /* TODO: Handle output modifiers */
1576 }
1577
1578 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1579 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1580 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1581 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1582 */
1583 void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG* arg) {
1584 SHADER_BUFFER* buffer = arg->buffer;
1585 DWORD dst_reg = arg->dst & WINED3DSP_REGNUM_MASK;
1586 char src0[50];
1587
1588 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src0);
1589 shader_addline(buffer, "DP3 TMP.y, T%u, %s;\n", dst_reg, src0);
1590
1591 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1592 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1593 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1594 */
1595 shader_addline(buffer, "RCP TMP.y, TMP.y;\n");
1596 shader_addline(buffer, "MUL TMP.x, TMP.x, TMP.y;\n");
1597 shader_addline(buffer, "MIN TMP.x, TMP.x, one.r;\n");
1598 shader_addline(buffer, "MAX result.depth, TMP.x, 0.0;\n");
1599 }
1600
1601 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1602 Vertex/Pixel shaders to ARB_vertex_program codes */
1603 void shader_hw_mnxn(SHADER_OPCODE_ARG* arg) {
1604
1605 int i;
1606 int nComponents = 0;
1607 SHADER_OPCODE_ARG tmpArg;
1608
1609 memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
1610
1611 /* Set constants for the temporary argument */
1612 tmpArg.shader = arg->shader;
1613 tmpArg.buffer = arg->buffer;
1614 tmpArg.src[0] = arg->src[0];
1615 tmpArg.src_addr[0] = arg->src_addr[0];
1616 tmpArg.src_addr[1] = arg->src_addr[1];
1617 tmpArg.reg_maps = arg->reg_maps;
1618
1619 switch(arg->opcode->opcode) {
1620 case WINED3DSIO_M4x4:
1621 nComponents = 4;
1622 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1623 break;
1624 case WINED3DSIO_M4x3:
1625 nComponents = 3;
1626 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
1627 break;
1628 case WINED3DSIO_M3x4:
1629 nComponents = 4;
1630 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1631 break;
1632 case WINED3DSIO_M3x3:
1633 nComponents = 3;
1634 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1635 break;
1636 case WINED3DSIO_M3x2:
1637 nComponents = 2;
1638 tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
1639 break;
1640 default:
1641 break;
1642 }
1643
1644 for (i = 0; i < nComponents; i++) {
1645 tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
1646 tmpArg.src[1] = arg->src[1]+i;
1647 vshader_hw_map2gl(&tmpArg);
1648 }
1649 }
1650
1651 void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG* arg) {
1652 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1653 SHADER_BUFFER* buffer = arg->buffer;
1654 DWORD dst = arg->dst;
1655 DWORD src = arg->src[0];
1656 DWORD swizzle = (src & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
1657
1658 char tmpLine[256];
1659
1660 strcpy(tmpLine, curOpcode->glname); /* Opcode */
1661 vshader_program_add_param(arg, dst, FALSE, tmpLine); /* Destination */
1662 strcat(tmpLine, ",");
1663 vshader_program_add_param(arg, src, TRUE, tmpLine);
1664 if ((WINED3DSP_NOSWIZZLE >> WINED3DSP_SWIZZLE_SHIFT) == swizzle) {
1665 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1666 * .w is used
1667 */
1668 strcat(tmpLine, ".w");
1669 }
1670
1671 shader_addline(buffer, "%s;\n", tmpLine);
1672 }
1673
1674 void shader_hw_nrm(SHADER_OPCODE_ARG* arg) {
1675 SHADER_BUFFER* buffer = arg->buffer;
1676 char dst_name[50];
1677 char src_name[50];
1678 char dst_wmask[20];
1679 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1680 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1681
1682 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1683 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1684
1685 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1686 shader_addline(buffer, "DP3 TMP, %s, %s;\n", src_name, src_name);
1687 shader_addline(buffer, "RSQ TMP, TMP.x;\n");
1688 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1689 shader_addline(buffer, "MUL%s %s%s, %s, TMP;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1690 src_name);
1691
1692 if (shift != 0)
1693 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1694 }
1695
1696 void shader_hw_sincos(SHADER_OPCODE_ARG* arg) {
1697 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1698 * must contain fixed constants. So we need a separate function to filter those constants and
1699 * can't use map2gl
1700 */
1701 SHADER_BUFFER* buffer = arg->buffer;
1702 char dst_name[50];
1703 char src_name[50];
1704 char dst_wmask[20];
1705 DWORD shift = (arg->dst & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
1706 BOOL sat = (arg->dst & WINED3DSP_DSTMOD_MASK) & WINED3DSPDM_SATURATE;
1707
1708 pshader_get_register_name(arg->shader, arg->dst, dst_name);
1709 shader_arb_get_write_mask(arg, arg->dst, dst_wmask);
1710
1711 pshader_gen_input_modifier_line(arg->shader, buffer, arg->src[0], 0, src_name);
1712 shader_addline(buffer, "SCS%s %s%s, %s;\n", sat ? "_SAT" : "", dst_name, dst_wmask,
1713 src_name);
1714
1715 if (shift != 0)
1716 pshader_gen_output_modifier_line(buffer, FALSE, dst_wmask, shift, dst_name);
1717
1718 }
1719
1720 /* TODO: merge with pixel shader */
1721 /* Map the opcode 1-to-1 to the GL code */
1722 void vshader_hw_map2gl(SHADER_OPCODE_ARG* arg) {
1723
1724 IWineD3DVertexShaderImpl *shader = (IWineD3DVertexShaderImpl*) arg->shader;
1725 CONST SHADER_OPCODE* curOpcode = arg->opcode;
1726 SHADER_BUFFER* buffer = arg->buffer;
1727 DWORD dst = arg->dst;
1728 DWORD* src = arg->src;
1729
1730 DWORD dst_regtype = shader_get_regtype(dst);
1731 char tmpLine[256];
1732 unsigned int i;
1733
1734 if ((curOpcode->opcode == WINED3DSIO_MOV && dst_regtype == WINED3DSPR_ADDR) || curOpcode->opcode == WINED3DSIO_MOVA) {
1735 memset(tmpLine, 0, sizeof(tmpLine));
1736 if(shader->rel_offset) {
1737 vshader_program_add_param(arg, src[0], TRUE, tmpLine);
1738 shader_addline(buffer, "ADD TMP.x, %s, helper_const.z;\n", tmpLine);
1739 shader_addline(buffer, "ARL A0.x, TMP.x;\n");
1740 } else {
1741 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1742 * with more than one component. Thus replicate the first source argument over all
1743 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1744 */
1745 DWORD parm = src[0] & ~(WINED3DVS_SWIZZLE_MASK);
1746 if((src[0] & WINED3DVS_X_W) == WINED3DVS_X_W) {
1747 parm |= WINED3DVS_X_W | WINED3DVS_Y_W | WINED3DVS_Z_W | WINED3DVS_W_W;
1748 } else if((src[0] & WINED3DVS_X_Z) == WINED3DVS_X_Z) {
1749 parm |= WINED3DVS_X_Z | WINED3DVS_Y_Z | WINED3DVS_Z_Z | WINED3DVS_W_Z;
1750 } else if((src[0] & WINED3DVS_X_Y) == WINED3DVS_X_Y) {
1751 parm |= WINED3DVS_X_Y | WINED3DVS_Y_Y | WINED3DVS_Z_Y | WINED3DVS_W_Y;
1752 } else if((src[0] & WINED3DVS_X_X) == WINED3DVS_X_X) {
1753 parm |= WINED3DVS_X_X | WINED3DVS_Y_X | WINED3DVS_Z_X | WINED3DVS_W_X;
1754 }
1755 vshader_program_add_param(arg, parm, TRUE, tmpLine);
1756 shader_addline(buffer, "ARL A0.x, %s;\n", tmpLine);
1757 }
1758 return;
1759 } else
1760 strcpy(tmpLine, curOpcode->glname);
1761
1762 if (curOpcode->num_params > 0) {
1763 vshader_program_add_param(arg, dst, FALSE, tmpLine);
1764 for (i = 1; i < curOpcode->num_params; ++i) {
1765 strcat(tmpLine, ",");
1766 vshader_program_add_param(arg, src[i-1], TRUE, tmpLine);
1767 }
1768 }
1769 shader_addline(buffer, "%s;\n", tmpLine);
1770 }
1771
1772 static GLuint create_arb_blt_vertex_program(WineD3D_GL_Info *gl_info) {
1773 GLuint program_id = 0;
1774 const char *blt_vprogram =
1775 "!!ARBvp1.0\n"
1776 "PARAM c[1] = { { 1, 0.5 } };\n"
1777 "MOV result.position, vertex.position;\n"
1778 "MOV result.color, c[0].x;\n"
1779 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1780 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1781 "END\n";
1782
1783 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1784 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, program_id));
1785 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_vprogram), blt_vprogram));
1786
1787 if (glGetError() == GL_INVALID_OPERATION) {
1788 GLint pos;
1789 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1790 FIXME("Vertex program error at position %d: %s\n", pos,
1791 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1792 }
1793
1794 return program_id;
1795 }
1796
1797 static GLuint create_arb_blt_fragment_program(WineD3D_GL_Info *gl_info) {
1798 GLuint program_id = 0;
1799 const char *blt_fprogram =
1800 "!!ARBfp1.0\n"
1801 "TEMP R0;\n"
1802 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1803 "MOV result.depth.z, R0.x;\n"
1804 "END\n";
1805
1806 GL_EXTCALL(glGenProgramsARB(1, &program_id));
1807 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, program_id));
1808 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(blt_fprogram), blt_fprogram));
1809
1810 if (glGetError() == GL_INVALID_OPERATION) {
1811 GLint pos;
1812 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
1813 FIXME("Fragment program error at position %d: %s\n", pos,
1814 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
1815 }
1816
1817 return program_id;
1818 }
1819
1820 static void shader_arb_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
1821 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1822 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1823 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1824
1825 if (useVS) {
1826 TRACE("Using vertex shader\n");
1827
1828 priv->current_vprogram_id = ((IWineD3DVertexShaderImpl *)This->stateBlock->vertexShader)->baseShader.prgId;
1829
1830 /* Bind the vertex program */
1831 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1832 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1833
1834 /* Enable OpenGL vertex programs */
1835 glEnable(GL_VERTEX_PROGRAM_ARB);
1836 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1837 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1838 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
1839 priv->current_vprogram_id = 0;
1840 glDisable(GL_VERTEX_PROGRAM_ARB);
1841 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1842 }
1843
1844 if (usePS) {
1845 TRACE("Using pixel shader\n");
1846
1847 priv->current_fprogram_id = ((IWineD3DPixelShaderImpl *)This->stateBlock->pixelShader)->baseShader.prgId;
1848
1849 /* Bind the fragment program */
1850 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1851 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1852
1853 /* Enable OpenGL fragment programs */
1854 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1855 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1856 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1857 } else {
1858 priv->current_fprogram_id = 0;
1859
1860 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM) && !priv->use_arbfp_fixed_func) {
1861 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1862 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1863 * replacement shader
1864 */
1865 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1866 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1867 }
1868 }
1869 }
1870
1871 static void shader_arb_select_depth_blt(IWineD3DDevice *iface) {
1872 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1873 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1874 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1875
1876 if (!priv->depth_blt_vprogram_id) priv->depth_blt_vprogram_id = create_arb_blt_vertex_program(gl_info);
1877 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->depth_blt_vprogram_id));
1878 glEnable(GL_VERTEX_PROGRAM_ARB);
1879
1880 if (!priv->depth_blt_fprogram_id) priv->depth_blt_fprogram_id = create_arb_blt_fragment_program(gl_info);
1881 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->depth_blt_fprogram_id));
1882 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1883 }
1884
1885 static void shader_arb_deselect_depth_blt(IWineD3DDevice *iface) {
1886 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1887 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1888 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1889
1890 if (priv->current_vprogram_id) {
1891 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id));
1892 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1893
1894 glEnable(GL_VERTEX_PROGRAM_ARB);
1895 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1896
1897 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This, priv->current_vprogram_id);
1898 } else {
1899 glDisable(GL_VERTEX_PROGRAM_ARB);
1900 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1901 }
1902
1903 if (priv->current_fprogram_id) {
1904 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id));
1905 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1906
1907 glEnable(GL_FRAGMENT_PROGRAM_ARB);
1908 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1909
1910 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This, priv->current_fprogram_id);
1911 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
1912 glDisable(GL_FRAGMENT_PROGRAM_ARB);
1913 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1914 }
1915 }
1916
1917 static void shader_arb_cleanup(IWineD3DDevice *iface) {
1918 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1919 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1920 if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) glDisable(GL_VERTEX_PROGRAM_ARB);
1921 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) glDisable(GL_FRAGMENT_PROGRAM_ARB);
1922 }
1923
1924 static void shader_arb_destroy(IWineD3DBaseShader *iface) {
1925 IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
1926 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *) This->baseShader.device)->adapter->gl_info;
1927
1928 ENTER_GL();
1929 GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId));
1930 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1931 LEAVE_GL();
1932 This->baseShader.prgId = 0;
1933 This->baseShader.is_compiled = FALSE;
1934 }
1935
1936 static HRESULT shader_arb_alloc(IWineD3DDevice *iface) {
1937 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1938 This->shader_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
1939 return WINED3D_OK;
1940 }
1941
1942 static void shader_arb_free(IWineD3DDevice *iface) {
1943 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
1944 WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
1945 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->shader_priv;
1946
1947 if(priv->depth_blt_vprogram_id) {
1948 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_vprogram_id));
1949 }
1950 if(priv->depth_blt_fprogram_id) {
1951 GL_EXTCALL(glDeleteProgramsARB(1, &priv->depth_blt_fprogram_id));
1952 }
1953
1954 HeapFree(GetProcessHeap(), 0, This->shader_priv);
1955 }
1956
1957 static BOOL shader_arb_dirty_const(IWineD3DDevice *iface) {
1958 return TRUE;
1959 }
1960
1961 static void arbfp_add_sRGB_correction(SHADER_BUFFER *buffer, const char *fragcolor, const char *tmp1,
1962 const char *tmp2, const char *tmp3, const char *tmp4) {
1963 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1964
1965 /* Calculate the > 0.0031308 case */
1966 shader_addline(buffer, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1, fragcolor);
1967 shader_addline(buffer, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1, fragcolor);
1968 shader_addline(buffer, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1, fragcolor);
1969 shader_addline(buffer, "MUL %s, %s, srgb_mul_hi;\n", tmp1, tmp1);
1970 shader_addline(buffer, "SUB %s, %s, srgb_sub_hi;\n", tmp1, tmp1);
1971 /* Calculate the < case */
1972 shader_addline(buffer, "MUL %s, srgb_mul_low, %s;\n", tmp2, fragcolor);
1973 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1974 shader_addline(buffer, "SLT %s, srgb_comparison, %s;\n", tmp3, fragcolor);
1975 shader_addline(buffer, "SGE %s, srgb_comparison, %s;\n", tmp4, fragcolor);
1976 /* Store the components > 0.0031308 in the destination */
1977 shader_addline(buffer, "MUL %s, %s, %s;\n", fragcolor, tmp1, tmp3);
1978 /* Add the components that are < 0.0031308 */
1979 shader_addline(buffer, "MAD result.color.xyz, %s, %s, %s;\n", tmp2, tmp4, fragcolor);
1980 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1981 }
1982
1983 static void shader_arb_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
1984 IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
1985 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
1986 CONST DWORD *function = This->baseShader.function;
1987 const char *fragcolor;
1988 WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
1989 local_constant* lconst;
1990
1991 /* Create the hw ARB shader */
1992 shader_addline(buffer, "!!ARBfp1.0\n");
1993
1994 shader_addline(buffer, "TEMP TMP;\n"); /* Used in matrix ops */
1995 shader_addline(buffer, "TEMP TMP2;\n"); /* Used in matrix ops */
1996 shader_addline(buffer, "TEMP TA;\n"); /* Used for modifiers */
1997 shader_addline(buffer, "TEMP TB;\n"); /* Used for modifiers */
1998 shader_addline(buffer, "TEMP TC;\n"); /* Used for modifiers */
1999 shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2000 shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2001 shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2002
2003 /* Base Declarations */
2004 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2005
2006 /* We need two variables for fog blending */
2007 shader_addline(buffer, "TEMP TMP_FOG;\n");
2008 if (This->baseShader.hex_version >= WINED3DPS_VERSION(2,0)) {
2009 shader_addline(buffer, "TEMP TMP_COLOR;\n");
2010 }
2011
2012 /* Base Shader Body */
2013 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2014
2015 /* calculate fog and blend it
2016 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
2017 * -1/(e-s) and e/(e-s) respectively.
2018 */
2019 shader_addline(buffer, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
2020
2021 if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
2022 fragcolor = "R0";
2023 } else {
2024 fragcolor = "TMP_COLOR";
2025 }
2026 if(This->srgb_enabled) {
2027 arbfp_add_sRGB_correction(buffer, fragcolor, "TMP", "TMP2", "TA", "TB");
2028 }
2029 if (This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
2030 shader_addline(buffer, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor);
2031 shader_addline(buffer, "MOV result.color.a, %s.a;\n", fragcolor);
2032 }
2033
2034 shader_addline(buffer, "END\n");
2035
2036 /* TODO: change to resource.glObjectHandle or something like that */
2037 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2038
2039 TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
2040 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, This->baseShader.prgId));
2041
2042 TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
2043 /* Create the program and check for errors */
2044 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2045 buffer->bsize, buffer->buffer));
2046
2047 if (glGetError() == GL_INVALID_OPERATION) {
2048 GLint errPos;
2049 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2050 FIXME("HW PixelShader Error at position %d: %s\n",
2051 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2052 This->baseShader.prgId = -1;
2053 }
2054
2055 /* Load immediate constants */
2056 if(!This->baseShader.load_local_constsF) {
2057 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2058 float *value = (float *) lconst->value;
2059 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, lconst->idx, value));
2060 checkGLcall("glProgramLocalParameter4fvARB");
2061 }
2062 }
2063 }
2064
2065 static void shader_arb_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
2066 IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
2067 shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
2068 CONST DWORD *function = This->baseShader.function;
2069 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
2070 WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
2071 local_constant* lconst;
2072
2073 /* Create the hw ARB shader */
2074 shader_addline(buffer, "!!ARBvp1.0\n");
2075 shader_addline(buffer, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This->rel_offset);
2076
2077 /* Mesa supports only 95 constants */
2078 if (GL_VEND(MESA) || GL_VEND(WINE))
2079 This->baseShader.limits.constant_float =
2080 min(95, This->baseShader.limits.constant_float);
2081
2082 shader_addline(buffer, "TEMP TMP;\n");
2083
2084 /* Base Declarations */
2085 shader_generate_arb_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
2086
2087 /* We need a constant to fixup the final position */
2088 shader_addline(buffer, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS);
2089
2090 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2091 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2092 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2093 * a replacement shader depend on the texcoord.w being set properly.
2094 *
2095 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2096 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2097 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2098 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2099 * this can eat a number of instructions, so skip it unless this cap is set as well
2100 */
2101 if(!GL_SUPPORT(NV_VERTEX_PROGRAM)) {
2102 shader_addline(buffer, "MOV result.color.secondary, -helper_const.wwwy;\n");
2103
2104 if((GLINFO_LOCATION).set_texcoord_w && !device->frag_pipe->ffp_proj_control) {
2105 int i;
2106 for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
2107 if(This->baseShader.reg_maps.texcoord_mask[i] != 0 &&
2108 This->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
2109 shader_addline(buffer, "MOV result.texcoord[%u].w, -helper_const.y;\n", i);
2110 }
2111 }
2112 }
2113 }
2114
2115 /* Base Shader Body */
2116 shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
2117
2118 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2119 if (!reg_maps->fog)
2120 shader_addline(buffer, "MOV result.fogcoord, TMP_OUT.z;\n");
2121
2122 /* Write the final position.
2123 *
2124 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2125 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2126 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2127 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2128 */
2129 shader_addline(buffer, "MUL TMP, posFixup, TMP_OUT.w;\n");
2130 shader_addline(buffer, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2131 shader_addline(buffer, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2132
2133 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2134 * and the glsl equivalent
2135 */
2136 shader_addline(buffer, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2137
2138 shader_addline(buffer, "MOV result.position, TMP_OUT;\n");
2139
2140 shader_addline(buffer, "END\n");
2141
2142 /* TODO: change to resource.glObjectHandle or something like that */
2143 GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
2144
2145 TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
2146 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
2147
2148 TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
2149 /* Create the program and check for errors */
2150 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
2151 buffer->bsize, buffer->buffer));
2152
2153 if (glGetError() == GL_INVALID_OPERATION) {
2154 GLint errPos;
2155 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
2156 FIXME("HW VertexShader Error at position %d: %s\n",
2157 errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2158 This->baseShader.prgId = -1;
2159 }
2160
2161 /* Load immediate constants */
2162 if(!This->baseShader.load_local_constsF) {
2163 LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
2164 float *value = (float *) lconst->value;
2165 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB, lconst->idx, value));
2166 }
2167 }
2168 }
2169
2170 static void shader_arb_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
2171 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2172 * then overwrite the shader specific ones
2173 */
2174 none_shader_backend.shader_get_caps(devtype, gl_info, pCaps);
2175
2176 if(GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
2177 pCaps->VertexShaderVersion = WINED3DVS_VERSION(1,1);
2178 TRACE_(d3d_caps)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2179 pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
2180 }
2181
2182 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM)) {
2183 pCaps->PixelShaderVersion = WINED3DPS_VERSION(1,4);
2184 pCaps->PixelShader1xMaxValue = 8.0;
2185 TRACE_(d3d_caps)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2186 }
2187 }
2188
2189 static BOOL shader_arb_conv_supported(WINED3DFORMAT fmt) {
2190 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt));
2191 switch(fmt) {
2192 case WINED3DFMT_V8U8:
2193 case WINED3DFMT_V16U16:
2194 case WINED3DFMT_X8L8V8U8:
2195 case WINED3DFMT_L6V5U5:
2196 case WINED3DFMT_Q8W8V8U8:
2197 case WINED3DFMT_ATI2N:
2198 TRACE("[OK]\n");
2199 return TRUE;
2200 default:
2201 TRACE("[FAILED\n");
2202 return FALSE;
2203 }
2204 }
2205
2206 const shader_backend_t arb_program_shader_backend = {
2207 shader_arb_select,
2208 shader_arb_select_depth_blt,
2209 shader_arb_deselect_depth_blt,
2210 shader_arb_load_constants,
2211 shader_arb_cleanup,
2212 shader_arb_color_correction,
2213 shader_arb_destroy,
2214 shader_arb_alloc,
2215 shader_arb_free,
2216 shader_arb_dirty_const,
2217 shader_arb_generate_pshader,
2218 shader_arb_generate_vshader,
2219 shader_arb_get_caps,
2220 shader_arb_conv_supported,
2221 };
2222
2223 /* ARB_fragment_program fixed function pipeline replacement definitions */
2224 #define ARB_FFP_CONST_TFACTOR 0
2225 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2226 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2227 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2228 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2229
2230 struct arbfp_ffp_desc
2231 {
2232 struct ffp_desc parent;
2233 GLuint shader;
2234 unsigned int num_textures_used;
2235 };
2236
2237 static void arbfp_enable(IWineD3DDevice *iface, BOOL enable) {
2238 if(enable) {
2239 glEnable(GL_FRAGMENT_PROGRAM_ARB);
2240 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2241 } else {
2242 glDisable(GL_FRAGMENT_PROGRAM_ARB);
2243 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2244 }
2245 }
2246
2247 static HRESULT arbfp_alloc(IWineD3DDevice *iface) {
2248 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2249 struct shader_arb_priv *priv;
2250 /* Share private data between the shader backend and the pipeline replacement, if both
2251 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2252 * if no pixel shader is bound or not
2253 */
2254 if(This->shader_backend == &arb_program_shader_backend) {
2255 This->fragment_priv = This->shader_priv;
2256 } else {
2257 This->fragment_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_arb_priv));
2258 if(!This->fragment_priv) return E_OUTOFMEMORY;
2259 }
2260 priv = (struct shader_arb_priv *) This->fragment_priv;
2261 priv->fragment_shaders = hash_table_create(ffp_program_key_hash, ffp_program_key_compare);
2262 priv->use_arbfp_fixed_func = TRUE;
2263 return WINED3D_OK;
2264 }
2265
2266 static void arbfp_free_ffpshader(void *value, void *gli) {
2267 WineD3D_GL_Info *gl_info = gli;
2268 struct arbfp_ffp_desc *entry_arb = value;
2269
2270 ENTER_GL();
2271 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb->shader));
2272 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2273 HeapFree(GetProcessHeap(), 0, entry_arb);
2274 LEAVE_GL();
2275 }
2276
2277 static void arbfp_free(IWineD3DDevice *iface) {
2278 IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *) iface;
2279 struct shader_arb_priv *priv = (struct shader_arb_priv *) This->fragment_priv;
2280
2281 hash_table_destroy(priv->fragment_shaders, arbfp_free_ffpshader, &This->adapter->gl_info);
2282 priv->use_arbfp_fixed_func = FALSE;
2283
2284 if(This->shader_backend != &arb_program_shader_backend) {
2285 HeapFree(GetProcessHeap(), 0, This->fragment_priv);
2286 }
2287 }
2288
2289 static void arbfp_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct fragment_caps *caps) {
2290 caps->TextureOpCaps = WINED3DTEXOPCAPS_DISABLE |
2291 WINED3DTEXOPCAPS_SELECTARG1 |
2292 WINED3DTEXOPCAPS_SELECTARG2 |
2293 WINED3DTEXOPCAPS_MODULATE4X |
2294 WINED3DTEXOPCAPS_MODULATE2X |
2295 WINED3DTEXOPCAPS_MODULATE |
2296 WINED3DTEXOPCAPS_ADDSIGNED2X |
2297 WINED3DTEXOPCAPS_ADDSIGNED |
2298 WINED3DTEXOPCAPS_ADD |
2299 WINED3DTEXOPCAPS_SUBTRACT |
2300 WINED3DTEXOPCAPS_ADDSMOOTH |
2301 WINED3DTEXOPCAPS_BLENDCURRENTALPHA |
2302 WINED3DTEXOPCAPS_BLENDFACTORALPHA |
2303 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA |
2304 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA |
2305 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM |
2306 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR |
2307 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA |
2308 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA |
2309 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR |
2310 WINED3DTEXOPCAPS_DOTPRODUCT3 |
2311 WINED3DTEXOPCAPS_MULTIPLYADD |
2312 WINED3DTEXOPCAPS_LERP |
2313 WINED3DTEXOPCAPS_BUMPENVMAP |
2314 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE;
2315
2316 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2317
2318 caps->MaxTextureBlendStages = 8;
2319 caps->MaxSimultaneousTextures = min(GL_LIMITS(fragment_samplers), 8);
2320
2321 caps->PrimitiveMiscCaps |= WINED3DPMISCCAPS_TSSARGTEMP;
2322 }
2323 #undef GLINFO_LOCATION
2324
2325 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2326 static void state_texfactor_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2327 float col[4];
2328 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2329
2330 /* Do not overwrite pixel shader constants if a pshader is in use */
2331 if(use_ps(device)) return;
2332
2333 D3DCOLORTOGLFLOAT4(stateblock->renderState[WINED3DRS_TEXTUREFACTOR], col);
2334 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col));
2335 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2336
2337 if(device->shader_backend == &arb_program_shader_backend) {
2338 device = stateblock->wineD3DDevice;
2339 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_TFACTOR] = 1;
2340 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_TFACTOR + 1);
2341 }
2342 }
2343
2344 static void state_arb_specularenable(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2345 float col[4];
2346 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2347
2348 /* Do not overwrite pixel shader constants if a pshader is in use */
2349 if(use_ps(device)) return;
2350
2351 if(stateblock->renderState[WINED3DRS_SPECULARENABLE]) {
2352 /* The specular color has no alpha */
2353 col[0] = 1.0; col[1] = 1.0;
2354 col[2] = 1.0; col[3] = 0.0;
2355 } else {
2356 col[0] = 0.0; col[1] = 0.0;
2357 col[2] = 0.0; col[3] = 0.0;
2358 }
2359 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col));
2360 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2361
2362 if(device->shader_backend == &arb_program_shader_backend) {
2363 device = stateblock->wineD3DDevice;
2364 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_SPECULAR_ENABLE] = 1;
2365 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_SPECULAR_ENABLE + 1);
2366 }
2367 }
2368
2369 static void set_bumpmat_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2370 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2371 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2372 float mat[2][2];
2373
2374 if(use_ps(device)) {
2375 if(stage != 0 &&
2376 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.bumpmat[stage]) {
2377 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2378 * anyway
2379 */
2380 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2381 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2382 }
2383 }
2384 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2385 return;
2386 }
2387
2388 mat[0][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT00]);
2389 mat[0][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT01]);
2390 mat[1][0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT10]);
2391 mat[1][1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVMAT11]);
2392
2393 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0]));
2394 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2395
2396 if(device->shader_backend == &arb_program_shader_backend) {
2397 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_BUMPMAT(stage)] = 1;
2398 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_BUMPMAT(stage) + 1);
2399 }
2400 }
2401
2402 static void tex_bumpenvlum_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2403 DWORD stage = (state - STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE;
2404 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2405 float param[4];
2406
2407 if(use_ps(device)) {
2408 if(stage != 0 &&
2409 ((IWineD3DPixelShaderImpl *) stateblock->pixelShader)->baseShader.reg_maps.luminanceparams[stage]) {
2410 /* The pixel shader has to know the luminance offset. Do a constants update if it
2411 * isn't scheduled anyway
2412 */
2413 if(!isStateDirty(context, STATE_PIXELSHADERCONSTANT)) {
2414 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2415 }
2416 }
2417 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2418 return;
2419 }
2420
2421 param[0] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLSCALE]);
2422 param[1] = *((float *) &stateblock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET]);
2423 param[2] = 0.0;
2424 param[3] = 0.0;
2425
2426 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param));
2427 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2428
2429 if(device->shader_backend == &arb_program_shader_backend) {
2430 device->activeContext->pshader_const_dirty[ARB_FFP_CONST_LUMINANCE(stage)] = 1;
2431 device->highest_dirty_ps_const = max(device->highest_dirty_ps_const, ARB_FFP_CONST_LUMINANCE(stage) + 1);
2432 }
2433 }
2434
2435 static const char *get_argreg(SHADER_BUFFER *buffer, DWORD argnum, unsigned int stage, DWORD arg) {
2436 const char *ret;
2437
2438 if(arg == ARG_UNUSED) return "unused"; /* This is the marker for unused registers */
2439
2440 switch(arg & WINED3DTA_SELECTMASK) {
2441 case WINED3DTA_DIFFUSE:
2442 ret = "fragment.color.primary"; break;
2443
2444 case WINED3DTA_CURRENT:
2445 if(stage == 0) ret = "fragment.color.primary";
2446 else ret = "ret";
2447 break;
2448
2449 case WINED3DTA_TEXTURE:
2450 switch(stage) {
2451 case 0: ret = "tex0"; break;
2452 case 1: ret = "tex1"; break;
2453 case 2: ret = "tex2"; break;
2454 case 3: ret = "tex3"; break;
2455 case 4: ret = "tex4"; break;
2456 case 5: ret = "tex5"; break;
2457 case 6: ret = "tex6"; break;
2458 case 7: ret = "tex7"; break;
2459 default: ret = "unknown texture";
2460 }
2461 break;
2462
2463 case WINED3DTA_TFACTOR:
2464 ret = "tfactor"; break;
2465
2466 case WINED3DTA_SPECULAR:
2467 ret = "fragment.color.secondary"; break;
2468
2469 case WINED3DTA_TEMP:
2470 ret = "tempreg"; break;
2471
2472 case WINED3DTA_CONSTANT:
2473 FIXME("Implement perstage constants\n");
2474 switch(stage) {
2475 case 0: ret = "const0"; break;
2476 case 1: ret = "const1"; break;
2477 case 2: ret = "const2"; break;
2478 case 3: ret = "const3"; break;
2479 case 4: ret = "const4"; break;
2480 case 5: ret = "const5"; break;
2481 case 6: ret = "const6"; break;
2482 case 7: ret = "const7"; break;
2483 }
2484 default:
2485 return "unknown";
2486 }
2487
2488 if(arg & WINED3DTA_COMPLEMENT) {
2489 shader_addline(buffer, "SUB arg%u, const.x, %s;\n", argnum, ret);
2490 if(argnum == 0) ret = "arg0";
2491 if(argnum == 1) ret = "arg1";
2492 if(argnum == 2) ret = "arg2";
2493 }
2494 if(arg & WINED3DTA_ALPHAREPLICATE) {
2495 shader_addline(buffer, "MOV arg%u, %s.a;\n", argnum, ret);
2496 if(argnum == 0) ret = "arg0";
2497 if(argnum == 1) ret = "arg1";
2498 if(argnum == 2) ret = "arg2";
2499 }
2500 return ret;
2501 }
2502
2503 static void gen_ffp_instr(SHADER_BUFFER *buffer, unsigned int stage, BOOL color, BOOL alpha,
2504 DWORD dst, DWORD op, DWORD dw_arg0, DWORD dw_arg1, DWORD dw_arg2) {
2505 const char *dstmask, *dstreg, *arg0, *arg1, *arg2;
2506 unsigned int mul = 1;
2507 BOOL mul_final_dest = FALSE;
2508
2509 if(color && alpha) dstmask = "";
2510 else if(color) dstmask = ".rgb";
2511 else dstmask = ".a";
2512
2513 if(dst == tempreg) dstreg = "tempreg";
2514 else dstreg = "ret";
2515
2516 arg0 = get_argreg(buffer, 0, stage, dw_arg0);
2517 arg1 = get_argreg(buffer, 1, stage, dw_arg1);
2518 arg2 = get_argreg(buffer, 2, stage, dw_arg2);
2519
2520 switch(op) {
2521 case WINED3DTOP_DISABLE:
2522 if(stage == 1) shader_addline(buffer, "MOV %s%s, fragment.color.primary;\n", dstreg, dstmask);
2523 break;
2524
2525 case WINED3DTOP_SELECTARG2:
2526 arg1 = arg2;
2527 case WINED3DTOP_SELECTARG1:
2528 shader_addline(buffer, "MOV %s%s, %s;\n", dstreg, dstmask, arg1);
2529 break;
2530
2531 case WINED3DTOP_MODULATE4X:
2532 mul = 2;
2533 case WINED3DTOP_MODULATE2X:
2534 mul *= 2;
2535 if(strcmp(dstreg, "result.color") == 0) {
2536 dstreg = "ret";
2537 mul_final_dest = TRUE;
2538 }
2539 case WINED3DTOP_MODULATE:
2540 shader_addline(buffer, "MUL %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2541 break;
2542
2543 case WINED3DTOP_ADDSIGNED2X:
2544 mul = 2;
2545 if(strcmp(dstreg, "result.color") == 0) {
2546 dstreg = "ret";
2547 mul_final_dest = TRUE;
2548 }
2549 case WINED3DTOP_ADDSIGNED:
2550 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2551 arg2 = "arg2";
2552 case WINED3DTOP_ADD:
2553 shader_addline(buffer, "ADD_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2554 break;
2555
2556 case WINED3DTOP_SUBTRACT:
2557 shader_addline(buffer, "SUB_SAT %s%s, %s, %s;\n", dstreg, dstmask, arg1, arg2);
2558 break;
2559
2560 case WINED3DTOP_ADDSMOOTH:
2561 shader_addline(buffer, "SUB arg1, const.x, %s;\n", arg1);
2562 shader_addline(buffer, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2563 break;
2564
2565 case WINED3DTOP_BLENDCURRENTALPHA:
2566 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_CURRENT);
2567 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2568 break;
2569 case WINED3DTOP_BLENDFACTORALPHA:
2570 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TFACTOR);
2571 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2572 break;
2573 case WINED3DTOP_BLENDTEXTUREALPHA:
2574 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2575 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2576 break;
2577 case WINED3DTOP_BLENDDIFFUSEALPHA:
2578 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_DIFFUSE);
2579 shader_addline(buffer, "LRP %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2580 break;
2581
2582 case WINED3DTOP_BLENDTEXTUREALPHAPM:
2583 arg0 = get_argreg(buffer, 0, stage, WINED3DTA_TEXTURE);
2584 shader_addline(buffer, "SUB arg0.a, const.x, %s.a;\n", arg0);
2585 shader_addline(buffer, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg, dstmask, arg2, arg1);
2586 break;
2587
2588 /* D3DTOP_PREMODULATE ???? */
2589
2590 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR:
2591 shader_addline(buffer, "SUB arg0.a, const.x, %s;\n", arg1);
2592 shader_addline(buffer, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg, dstmask, arg2, arg1);
2593 break;
2594 case WINED3DTOP_MODULATEALPHA_ADDCOLOR:
2595 shader_addline(buffer, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg1);
2596 break;
2597 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA:
2598 shader_addline(buffer, "SUB arg0, const.x, %s;\n", arg1);
2599 shader_addline(buffer, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg, dstmask, arg2, arg1);
2600 break;
2601 case WINED3DTOP_MODULATECOLOR_ADDALPHA:
2602 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg, dstmask, arg1, arg2, arg1);
2603 break;
2604
2605 case WINED3DTOP_DOTPRODUCT3:
2606 mul = 4;
2607 if(strcmp(dstreg, "result.color") == 0) {
2608 dstreg = "ret";
2609 mul_final_dest = TRUE;
2610 }
2611 shader_addline(buffer, "SUB arg1, %s, const.w;\n", arg1);
2612 shader_addline(buffer, "SUB arg2, %s, const.w;\n", arg2);
2613 shader_addline(buffer, "DP3_SAT %s%s, arg1, arg2;\n", dstreg, dstmask);
2614 break;
2615
2616 case WINED3DTOP_MULTIPLYADD:
2617 shader_addline(buffer, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg, dstmask, arg1, arg2, arg0);
2618 break;
2619
2620 case WINED3DTOP_LERP:
2621 /* The msdn is not quite right here */
2622 shader_addline(buffer, "LRP %s%s, %s, %s, %s;\n", dstreg, dstmask, arg0, arg1, arg2);
2623 break;
2624
2625 case WINED3DTOP_BUMPENVMAP:
2626 case WINED3DTOP_BUMPENVMAPLUMINANCE:
2627 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2628 break;
2629
2630 default:
2631 FIXME("Unhandled texture op %08x\n", op);
2632 }
2633
2634 if(mul == 2) {
2635 shader_addline(buffer, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2636 } else if(mul == 4) {
2637 shader_addline(buffer, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest ? "result.color" : dstreg, dstmask, dstreg);
2638 }
2639 }
2640
2641 /* The stateblock is passed for GLINFO_LOCATION */
2642 static GLuint gen_arbfp_ffp_shader(struct ffp_settings *settings, IWineD3DStateBlockImpl *stateblock) {
2643 unsigned int stage;
2644 SHADER_BUFFER buffer;
2645 BOOL tex_read[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2646 BOOL bump_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2647 BOOL luminance_used[MAX_TEXTURES] = {FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE};
2648 const char *textype;
2649 const char *instr, *sat;
2650 char colorcor_dst[8];
2651 GLuint ret;
2652 DWORD arg0, arg1, arg2;
2653 BOOL tempreg_used = FALSE, tfactor_used = FALSE;
2654 BOOL op_equal;
2655 const char *final_combiner_src = "ret";
2656
2657 /* Find out which textures are read */
2658 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2659 if(settings->op[stage].cop == WINED3DTOP_DISABLE) break;
2660 arg0 = settings->op[stage].carg0 & WINED3DTA_SELECTMASK;
2661 arg1 = settings->op[stage].carg1 & WINED3DTA_SELECTMASK;
2662 arg2 = settings->op[stage].carg2 & WINED3DTA_SELECTMASK;
2663 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2664 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2665 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2666
2667 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHA) tex_read[stage] = TRUE;
2668 if(settings->op[stage].cop == WINED3DTOP_BLENDTEXTUREALPHAPM) tex_read[stage] = TRUE;
2669 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP) {
2670 bump_used[stage] = TRUE;
2671 tex_read[stage] = TRUE;
2672 }
2673 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2674 bump_used[stage] = TRUE;
2675 tex_read[stage] = TRUE;
2676 luminance_used[stage] = TRUE;
2677 } else if(settings->op[stage].cop == WINED3DTOP_BLENDFACTORALPHA) {
2678 tfactor_used = TRUE;
2679 }
2680
2681 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2682 tfactor_used = TRUE;
2683 }
2684
2685 if(settings->op[stage].dst == tempreg) tempreg_used = TRUE;
2686 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2687 tempreg_used = TRUE;
2688 }
2689
2690 if(settings->op[stage].aop == WINED3DTOP_DISABLE) continue;
2691 arg0 = settings->op[stage].aarg0 & WINED3DTA_SELECTMASK;
2692 arg1 = settings->op[stage].aarg1 & WINED3DTA_SELECTMASK;
2693 arg2 = settings->op[stage].aarg2 & WINED3DTA_SELECTMASK;
2694 if(arg0 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2695 if(arg1 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2696 if(arg2 == WINED3DTA_TEXTURE) tex_read[stage] = TRUE;
2697
2698 if(arg0 == WINED3DTA_TEMP || arg1 == WINED3DTA_TEMP || arg2 == WINED3DTA_TEMP) {
2699 tempreg_used = TRUE;
2700 }
2701 if(arg0 == WINED3DTA_TFACTOR || arg1 == WINED3DTA_TFACTOR || arg2 == WINED3DTA_TFACTOR) {
2702 tfactor_used = TRUE;
2703 }
2704 }
2705
2706 /* Shader header */
2707 buffer.bsize = 0;
2708 buffer.lineNo = 0;
2709 buffer.newline = TRUE;
2710 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
2711
2712 shader_addline(&buffer, "!!ARBfp1.0\n");
2713
2714 switch(settings->fog) {
2715 case FOG_OFF: break;
2716 case FOG_LINEAR: shader_addline(&buffer, "OPTION ARB_fog_linear;\n"); break;
2717 case FOG_EXP: shader_addline(&buffer, "OPTION ARB_fog_exp;\n"); break;
2718 case FOG_EXP2: shader_addline(&buffer, "OPTION ARB_fog_exp2;\n"); break;
2719 default: FIXME("Unexpected fog setting %d\n", settings->fog);
2720 }
2721
2722 shader_addline(&buffer, "PARAM const = {1, 2, 4, 0.5};\n");
2723 shader_addline(&buffer, "TEMP ret;\n");
2724 if(tempreg_used || settings->sRGB_write) shader_addline(&buffer, "TEMP tempreg;\n");
2725 shader_addline(&buffer, "TEMP arg0;\n");
2726 shader_addline(&buffer, "TEMP arg1;\n");
2727 shader_addline(&buffer, "TEMP arg2;\n");
2728 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2729 if(!tex_read[stage]) continue;
2730 shader_addline(&buffer, "TEMP tex%u;\n", stage);
2731 if(!bump_used[stage]) continue;
2732 shader_addline(&buffer, "PARAM bumpmat%u = program.env[%u];\n", stage, ARB_FFP_CONST_BUMPMAT(stage));
2733 if(!luminance_used[stage]) continue;
2734 shader_addline(&buffer, "PARAM luminance%u = program.env[%u];\n", stage, ARB_FFP_CONST_LUMINANCE(stage));
2735 }
2736 if(tfactor_used) {
2737 shader_addline(&buffer, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR);
2738 }
2739 shader_addline(&buffer, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE);
2740
2741 if(settings->sRGB_write) {
2742 shader_addline(&buffer, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2743 srgb_mul_low, srgb_mul_low, srgb_mul_low);
2744 shader_addline(&buffer, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2745 srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
2746 shader_addline(&buffer, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2747 srgb_pow, srgb_pow, srgb_pow);
2748 shader_addline(&buffer, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2749 srgb_mul_high, srgb_mul_high, srgb_mul_high);
2750 shader_addline(&buffer, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2751 srgb_sub_high, srgb_sub_high, srgb_sub_high);
2752 }
2753
2754 /* Generate texture sampling instructions) */
2755 for(stage = 0; stage < MAX_TEXTURES && settings->op[stage].cop != WINED3DTOP_DISABLE; stage++) {
2756 if(!tex_read[stage]) continue;
2757
2758 switch(settings->op[stage].tex_type) {
2759 case tex_1d: textype = "1D"; break;
2760 case tex_2d: textype = "2D"; break;
2761 case tex_3d: textype = "3D"; break;
2762 case tex_cube: textype = "CUBE"; break;
2763 case tex_rect: textype = "RECT"; break;
2764 default: textype = "unexpected_textype"; break;
2765 }
2766
2767 if(settings->op[stage].cop == WINED3DTOP_BUMPENVMAP ||
2768 settings->op[stage].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2769 sat = "";
2770 } else {
2771 sat = "_SAT";
2772 }
2773
2774 if(settings->op[stage].projected == proj_none) {
2775 instr = "TEX";
2776 } else if(settings->op[stage].projected == proj_count4 ||
2777 settings->op[stage].projected == proj_count3) {
2778 instr = "TXP";
2779 } else {
2780 FIXME("Unexpected projection mode %d\n", settings->op[stage].projected);
2781 instr = "TXP";
2782 }
2783
2784 if(stage > 0 &&
2785 (settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAP ||
2786 settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE)) {
2787 shader_addline(&buffer, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage - 1);
2788 shader_addline(&buffer, "DP3 ret.r, arg1, tex%u;\n", stage - 1);
2789 shader_addline(&buffer, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage - 1);
2790 shader_addline(&buffer, "DP3 ret.g, arg1, tex%u;\n", stage - 1);
2791
2792 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2793 * so multiply the displacement with the dividing parameter before passing it to TXP
2794 */
2795 if (settings->op[stage].projected != proj_none) {
2796 if(settings->op[stage].projected == proj_count4) {
2797 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].a;\n", stage);
2798 shader_addline(&buffer, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage, stage);
2799 } else {
2800 shader_addline(&buffer, "MOV ret.a, fragment.texcoord[%u].b;\n", stage);
2801 shader_addline(&buffer, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage, stage);
2802 }
2803 } else {
2804 shader_addline(&buffer, "ADD ret, ret, fragment.texcoord[%u];\n", stage);
2805 }
2806
2807 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2808 instr, sat, stage, stage, textype);
2809 if(settings->op[stage - 1].cop == WINED3DTOP_BUMPENVMAPLUMINANCE) {
2810 shader_addline(&buffer, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2811 stage - 1, stage - 1, stage - 1);
2812 shader_addline(&buffer, "MUL tex%u, tex%u, ret.r;\n", stage, stage);
2813 }
2814 } else if(settings->op[stage].projected == proj_count3) {
2815 shader_addline(&buffer, "MOV ret, fragment.texcoord[%u];\n", stage);
2816 shader_addline(&buffer, "MOV ret.a, ret.b;\n");
2817 shader_addline(&buffer, "%s%s tex%u, ret, texture[%u], %s;\n",
2818 instr, sat, stage, stage, textype);
2819 } else {
2820 shader_addline(&buffer, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2821 instr, sat, stage, stage, stage, textype);
2822 }
2823
2824 sprintf(colorcor_dst, "tex%u", stage);
2825 gen_color_correction(&buffer, colorcor_dst, ".rgba", "const.x", "const.y",
2826 settings->op[stage].color_correction, &GLINFO_LOCATION);
2827 }
2828
2829 /* Generate the main shader */
2830 for(stage = 0; stage < MAX_TEXTURES; stage++) {
2831 if(settings->op[stage].cop == WINED3DTOP_DISABLE) {
2832 if(stage == 0) {
2833 final_combiner_src = "fragment.color.primary";
2834 }
2835 break;
2836 }
2837
2838 if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2839 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2840 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg1;
2841 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG1 &&
2842 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2843 op_equal = settings->op[stage].carg1 == settings->op[stage].aarg2;
2844 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2845 settings->op[stage].aop == WINED3DTOP_SELECTARG1) {
2846 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg1;
2847 } else if(settings->op[stage].cop == WINED3DTOP_SELECTARG2 &&
2848 settings->op[stage].aop == WINED3DTOP_SELECTARG2) {
2849 op_equal = settings->op[stage].carg2 == settings->op[stage].aarg2;
2850 } else {
2851 op_equal = settings->op[stage].aop == settings->op[stage].cop &&
2852 settings->op[stage].carg0 == settings->op[stage].aarg0 &&
2853 settings->op[stage].carg1 == settings->op[stage].aarg1 &&
2854 settings->op[stage].carg2 == settings->op[stage].aarg2;
2855 }
2856
2857 if(settings->op[stage].aop == WINED3DTOP_DISABLE) {
2858 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2859 settings->op[stage].cop, settings->op[stage].carg0,
2860 settings->op[stage].carg1, settings->op[stage].carg2);
2861 if(stage == 0) {
2862 shader_addline(&buffer, "MOV ret.a, fragment.color.primary.a;\n");
2863 }
2864 } else if(op_equal) {
2865 gen_ffp_instr(&buffer, stage, TRUE, TRUE, settings->op[stage].dst,
2866 settings->op[stage].cop, settings->op[stage].carg0,
2867 settings->op[stage].carg1, settings->op[stage].carg2);
2868 } else {
2869 gen_ffp_instr(&buffer, stage, TRUE, FALSE, settings->op[stage].dst,
2870 settings->op[stage].cop, settings->op[stage].carg0,
2871 settings->op[stage].carg1, settings->op[stage].carg2);
2872 gen_ffp_instr(&buffer, stage, FALSE, TRUE, settings->op[stage].dst,
2873 settings->op[stage].aop, settings->op[stage].aarg0,
2874 settings->op[stage].aarg1, settings->op[stage].aarg2);
2875 }
2876 }
2877
2878 if(settings->sRGB_write) {
2879 shader_addline(&buffer, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2880 arbfp_add_sRGB_correction(&buffer, "ret", "arg0", "arg1", "arg2", "tempreg");
2881 shader_addline(&buffer, "MOV result.color.a, ret.a;\n");
2882 } else {
2883 shader_addline(&buffer, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src);
2884 }
2885
2886 /* Footer */
2887 shader_addline(&buffer, "END\n");
2888
2889 /* Generate the shader */
2890 GL_EXTCALL(glGenProgramsARB(1, &ret));
2891 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, ret));
2892 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
2893
2894 if (glGetError() == GL_INVALID_OPERATION) {
2895 GLint pos;
2896 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
2897 FIXME("Vertex program error at position %d: %s\n", pos,
2898 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
2899 }
2900 HeapFree(GetProcessHeap(), 0, buffer.buffer);
2901 return ret;
2902 }
2903
2904 static void fragment_prog_arbfp(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2905 IWineD3DDeviceImpl *device = stateblock->wineD3DDevice;
2906 struct shader_arb_priv *priv = (struct shader_arb_priv *) device->fragment_priv;
2907 BOOL use_pshader = use_ps(device);
2908 BOOL use_vshader = use_vs(device);
2909 struct ffp_settings settings;
2910 struct arbfp_ffp_desc *desc;
2911 unsigned int i;
2912
2913 if(isStateDirty(context, STATE_RENDER(WINED3DRS_FOGENABLE))) {
2914 if(use_pshader) {
2915 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
2916 } else if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2917 /* Reload fixed function constants since they collide with the pixel shader constants */
2918 for(i = 0; i < MAX_TEXTURES; i++) {
2919 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2920 }
2921 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2922 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2923 }
2924 return;
2925 }
2926
2927 if(use_pshader) {
2928 IWineD3DPixelShader_CompileShader(stateblock->pixelShader);
2929 } else {
2930 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2931 gen_ffp_op(stateblock, &settings, FALSE);
2932 desc = (struct arbfp_ffp_desc *) find_ffp_shader(priv->fragment_shaders, &settings);
2933 if(!desc) {
2934 desc = HeapAlloc(GetProcessHeap(), 0, sizeof(*desc));
2935 if(!desc) {
2936 ERR("Out of memory\n");
2937 return;
2938 }
2939 desc->num_textures_used = 0;
2940 for(i = 0; i < GL_LIMITS(texture_stages); i++) {
2941 if(settings.op[i].cop == WINED3DTOP_DISABLE) break;
2942 desc->num_textures_used = i;
2943 }
2944
2945 memcpy(&desc->parent.settings, &settings, sizeof(settings));
2946 desc->shader = gen_arbfp_ffp_shader(&settings, stateblock);
2947 add_ffp_shader(priv->fragment_shaders, &desc->parent);
2948 TRACE("Allocated fixed function replacement shader descriptor %p\n", desc);
2949 }
2950
2951 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
2952 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
2953 * deactivate it.
2954 */
2955 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader));
2956 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
2957
2958 if(device->shader_backend == &arb_program_shader_backend && context->last_was_pshader) {
2959 /* Reload fixed function constants since they collide with the pixel shader constants */
2960 for(i = 0; i < MAX_TEXTURES; i++) {
2961 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i, WINED3DTSS_BUMPENVMAT00), stateblock, context);
2962 }
2963 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR), stateblock, context);
2964 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE), stateblock, context);
2965 }
2966 }
2967
2968 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
2969 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
2970 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
2971 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
2972 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
2973 *
2974 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
2975 * to be compiled before activating them(needs some cleanups in the shader backend interface)
2976 */
2977 if(!isStateDirty(context, device->StateTable[STATE_VSHADER].representative)) {
2978 device->shader_backend->shader_select((IWineD3DDevice *)stateblock->wineD3DDevice, use_pshader, use_vshader);
2979
2980 if (!isStateDirty(context, STATE_VERTEXSHADERCONSTANT) && (use_vshader || use_pshader)) {
2981 device->StateTable[STATE_VERTEXSHADERCONSTANT].apply(STATE_VERTEXSHADERCONSTANT, stateblock, context);
2982 }
2983 }
2984 if(use_pshader) {
2985 device->StateTable[STATE_PIXELSHADERCONSTANT].apply(STATE_PIXELSHADERCONSTANT, stateblock, context);
2986 }
2987 }
2988
2989 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
2990 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
2991 * state table, so we need to handle that with a forwarding function. The other invisible side effect
2992 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
2993 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
2994 */
2995 static void state_arbfp_fog(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
2996 if(!isStateDirty(context, STATE_PIXELSHADER)) {
2997 fragment_prog_arbfp(state, stateblock, context);
2998 }
2999 }
3000
3001 static void textransform(DWORD state, IWineD3DStateBlockImpl *stateblock, WineD3DContext *context) {
3002 if(!isStateDirty(context, STATE_PIXELSHADER)) {
3003 fragment_prog_arbfp(state, stateblock, context);
3004 }
3005 }
3006
3007 #undef GLINFO_LOCATION
3008
3009 static const struct StateEntryTemplate arbfp_fragmentstate_template[] = {
3010 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR), state_texfactor_arbfp }, 0 },
3011 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3012 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3013 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3014 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3015 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3016 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3017 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3018 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3019 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3020 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3021 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3022 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3023 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3024 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3025 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3026 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3027 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3028 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3029 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3030 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3031 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3032 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3033 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3034 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3035 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3036 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3037 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3038 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3039 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3040 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3041 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3042 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3043 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3044 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3045 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3046 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3047 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3048 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3049 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3050 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3051 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3052 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3053 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3054 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3055 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3056 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3057 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3058 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3059 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3060 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3061 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3062 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3063 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3064 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3065 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3066 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3067 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3068 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3069 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3070 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3071 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3072 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3073 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3074 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3075 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3076 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3077 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3078 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3079 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3080 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3081 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3082 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3083 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3084 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3085 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3086 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3087 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3088 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3089 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3090 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3091 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3092 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3093 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3094 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3095 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3096 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3097 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3098 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3099 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3100 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3101 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3102 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3103 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3104 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3105 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3106 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3107 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3108 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3109 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3110 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3111 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3112 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3113 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3114 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3115 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3116 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3117 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3118 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3119 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3120 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3121 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3122 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3123 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3124 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3125 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3126 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3127 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3128 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00), set_bumpmat_arbfp }, 0 },
3129 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3130 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE), tex_bumpenvlum_arbfp }, 0 },
3131 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim }, 0 },
3132 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim }, 0 },
3133 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim }, 0 },
3134 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim }, 0 },
3135 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim }, 0 },
3136 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim }, 0 },
3137 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim }, 0 },
3138 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim }, 0 },
3139 { STATE_PIXELSHADER, { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3140 { STATE_RENDER(WINED3DRS_FOGENABLE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3141 { STATE_RENDER(WINED3DRS_FOGTABLEMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3142 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE), { STATE_RENDER(WINED3DRS_FOGENABLE), state_arbfp_fog }, 0 },
3143 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE), { STATE_PIXELSHADER, fragment_prog_arbfp }, 0 },
3144 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3145 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3146 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3147 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3148 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3149 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3150 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3151 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS), textransform }, 0 },
3152 { STATE_RENDER(WINED3DRS_SPECULARENABLE), { STATE_RENDER(WINED3DRS_SPECULARENABLE), state_arb_specularenable}, 0 },
3153 {0 /* Terminate */, { 0, 0 }, 0 },
3154 };
3155
3156 const struct fragment_pipeline arbfp_fragment_pipeline = {
3157 arbfp_enable,
3158 arbfp_get_caps,
3159 arbfp_alloc,
3160 arbfp_free,
3161 shader_arb_conv_supported,
3162 arbfp_fragmentstate_template,
3163 TRUE /* We can disable projected textures */
3164 };
3165
3166 #define GLINFO_LOCATION device->adapter->gl_info
3167
3168 struct arbfp_blit_priv {
3169 GLenum yuy2_rect_shader, yuy2_2d_shader;
3170 GLenum uyvy_rect_shader, uyvy_2d_shader;
3171 GLenum yv12_rect_shader, yv12_2d_shader;
3172 };
3173
3174 static HRESULT arbfp_blit_alloc(IWineD3DDevice *iface) {
3175 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3176 device->blit_priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct arbfp_blit_priv));
3177 if(!device->blit_priv) {
3178 ERR("Out of memory\n");
3179 return E_OUTOFMEMORY;
3180 }
3181 return WINED3D_OK;
3182 }
3183 static void arbfp_blit_free(IWineD3DDevice *iface) {
3184 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3185 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3186
3187 ENTER_GL();
3188 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_rect_shader));
3189 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yuy2_2d_shader));
3190 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_rect_shader));
3191 GL_EXTCALL(glDeleteProgramsARB(1, &priv->uyvy_2d_shader));
3192 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_rect_shader));
3193 GL_EXTCALL(glDeleteProgramsARB(1, &priv->yv12_2d_shader));
3194 checkGLcall("Delete yuv programs\n");
3195 LEAVE_GL();
3196 }
3197
3198 BOOL gen_planar_yuv_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3199 char chroma;
3200 const char *tex, *texinstr;
3201
3202 if(fmt == WINED3DFMT_UYVY) {
3203 chroma = 'r';
3204 *luminance = 'a';
3205 } else {
3206 chroma = 'a';
3207 *luminance = 'r';
3208 }
3209 switch(textype) {
3210 case GL_TEXTURE_2D: tex = "2D"; texinstr = "TXP"; break;
3211 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; texinstr = "TEX"; break;
3212 default:
3213 /* This is more tricky than just replacing the texture type - we have to navigate
3214 * properly in the texture to find the correct chroma values
3215 */
3216 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3217 return FALSE;
3218 }
3219
3220 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3221 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3222 * filtering when we sample the texture.
3223 *
3224 * These are the rules for reading the chroma:
3225 *
3226 * Even pixel: Cr
3227 * Even pixel: U
3228 * Odd pixel: V
3229 *
3230 * So we have to get the sampling x position in non-normalized coordinates in integers
3231 */
3232 if(textype != GL_TEXTURE_RECTANGLE_ARB) {
3233 shader_addline(buffer, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3234 shader_addline(buffer, "MOV texcrd.a, size.x;\n");
3235 } else {
3236 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3237 }
3238 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3239 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3240 * 0.5, so add 0.5.
3241 */
3242 shader_addline(buffer, "FLR texcrd.x, texcrd.x;\n");
3243 shader_addline(buffer, "ADD texcrd.x, texcrd.x, coef.y;\n");
3244
3245 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3246 * even and odd pixels respectively
3247 */
3248 shader_addline(buffer, "MUL texcrd2, texcrd, coef.y;\n");
3249 shader_addline(buffer, "FRC texcrd2, texcrd2;\n");
3250
3251 /* Sample Pixel 1 */
3252 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3253
3254 /* Put the value into either of the chroma values */
3255 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3256 shader_addline(buffer, "MUL chroma.r, luminance.%c, temp.x;\n", chroma);
3257 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3258 shader_addline(buffer, "MUL chroma.g, luminance.%c, temp.x;\n", chroma);
3259
3260 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3261 * the pixel right to the current one. Otherwise, sample the left pixel.
3262 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3263 */
3264 shader_addline(buffer, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3265 shader_addline(buffer, "ADD texcrd.x, texcrd, temp.x;\n");
3266 shader_addline(buffer, "%s luminance, texcrd, texture[0], %s;\n", texinstr, tex);
3267
3268 /* Put the value into the other chroma */
3269 shader_addline(buffer, "SGE temp.x, texcrd2.x, coef.y;\n");
3270 shader_addline(buffer, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma);
3271 shader_addline(buffer, "SLT temp.x, texcrd2.x, coef.y;\n");
3272 shader_addline(buffer, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma);
3273
3274 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3275 * the current one and lerp the two U and V values
3276 */
3277
3278 /* This gives the correctly filtered luminance value */
3279 shader_addline(buffer, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex);
3280
3281 return TRUE;
3282 }
3283
3284 BOOL gen_yv12_read(SHADER_BUFFER *buffer, WINED3DFORMAT fmt, GLenum textype, char *luminance) {
3285 const char *tex;
3286
3287 switch(textype) {
3288 case GL_TEXTURE_2D: tex = "2D"; break;
3289 case GL_TEXTURE_RECTANGLE_ARB: tex = "RECT"; break;
3290 default:
3291 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3292 return FALSE;
3293 }
3294
3295 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3296 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3297 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3298 * pitch of the luminance plane, the packing into the gl texture is a bit
3299 * unfortunate. If the whole texture is interpreted as luminance data it looks
3300 * approximately like this:
3301 *
3302 * +----------------------------------+----
3303 * | |
3304 * | |
3305 * | |
3306 * | |
3307 * | | 2
3308 * | LUMINANCE | -
3309 * | | 3
3310 * | |
3311 * | |
3312 * | |
3313 * | |
3314 * +----------------+-----------------+----
3315 * | | |
3316 * | U even rows | U odd rows |
3317 * | | | 1
3318 * +----------------+------------------ -
3319 * | | | 3
3320 * | V even rows | V odd rows |
3321 * | | |
3322 * +----------------+-----------------+----
3323 * | | |
3324 * | 0.5 | 0.5 |
3325 *
3326 * So it appears as if there are 4 chroma images, but in fact the odd rows
3327 * in the chroma images are in the same row as the even ones. So its is
3328 * kinda tricky to read
3329 *
3330 * When reading from rectangle textures, keep in mind that the input y coordinates
3331 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3332 */
3333 shader_addline(buffer, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3334 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3335
3336 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3337 /* the chroma planes have only half the width */
3338 shader_addline(buffer, "MUL texcrd.x, texcrd.x, coef.y;\n");
3339
3340 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3341 * the coordinate. Also read the right side of the image when reading odd lines
3342 *
3343 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3344 * bleeding
3345 */
3346 if(textype == GL_TEXTURE_2D) {
3347
3348 shader_addline(buffer, "RCP chroma.w, size.y;\n");
3349
3350 shader_addline(buffer, "MUL texcrd2.y, texcrd.y, size.y;\n");
3351
3352 shader_addline(buffer, "FLR texcrd2.y, texcrd2.y;\n");
3353 shader_addline(buffer, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3354
3355 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3356 shader_addline(buffer, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3357 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3358 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3359 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3360
3361 /* clamp, keep the half pixel origin in mind */
3362 shader_addline(buffer, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3363 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3364 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3365 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3366 } else {
3367 /* Read from [size - size+size/4] */
3368 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3369 shader_addline(buffer, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3370
3371 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3372 shader_addline(buffer, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3373 shader_addline(buffer, "FRC texcrd2.x, texcrd2.x;\n");
3374 shader_addline(buffer, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3375 shader_addline(buffer, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3376 shader_addline(buffer, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3377
3378 /* Make sure to read exactly from the pixel center */
3379 shader_addline(buffer, "FLR texcrd.y, texcrd.y;\n");
3380 shader_addline(buffer, "ADD texcrd.y, texcrd.y, coef.y;\n");
3381
3382 /* Clamp */
3383 shader_addline(buffer, "MAD temp.y, size.y, coef.w, size.y;\n");
3384 shader_addline(buffer, "ADD temp.y, temp.y, -coef.y;\n");
3385 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3386 shader_addline(buffer, "ADD temp.y, size.y, -coef.y;\n");
3387 shader_addline(buffer, "MAX texcrd.y, temp.y, texcrd.y;\n");
3388 }
3389 /* Read the texture, put the result into the output register */
3390 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3391 shader_addline(buffer, "MOV chroma.r, temp.a;\n");
3392
3393 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3394 * No need to clamp because we're just reusing the already clamped value from above
3395 */
3396 if(textype == GL_TEXTURE_2D) {
3397 shader_addline(buffer, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3398 } else {
3399 shader_addline(buffer, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3400 }
3401 shader_addline(buffer, "TEX temp, texcrd, texture[0], %s;\n", tex);
3402 shader_addline(buffer, "MOV chroma.g, temp.a;\n");
3403
3404 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3405 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3406 * values due to filtering
3407 */
3408 shader_addline(buffer, "MOV texcrd, fragment.texcoord[0];\n");
3409 if(textype == GL_TEXTURE_2D) {
3410 /* Multiply the y coordinate by 2/3 and clamp it */
3411 shader_addline(buffer, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3412 shader_addline(buffer, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3413 shader_addline(buffer, "MIN texcrd.y, temp.y, texcrd.y;\n");
3414 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3415 } else {
3416 /* Reading from texture_rectangles is pretty streightforward, just use the unmodified
3417 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3418 * is bigger
3419 */
3420 shader_addline(buffer, "ADD temp.x, size.y, -coef.y;\n");
3421 shader_addline(buffer, "MIN texcrd.y, texcrd.y, size.x;\n");
3422 shader_addline(buffer, "TEX luminance, texcrd, texture[0], %s;\n", tex);
3423 }
3424 *luminance = 'a';
3425
3426 return TRUE;
3427 }
3428
3429 GLuint gen_yuv_shader(IWineD3DDeviceImpl *device, WINED3DFORMAT fmt, GLenum textype) {
3430 GLenum shader;
3431 SHADER_BUFFER buffer;
3432 char luminance_component;
3433 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3434
3435 /* Shader header */
3436 buffer.bsize = 0;
3437 buffer.lineNo = 0;
3438 buffer.newline = TRUE;
3439 buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
3440
3441 GL_EXTCALL(glGenProgramsARB(1, &shader));
3442 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3443 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3444 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3445 if(!shader) return 0;
3446
3447 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3448 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3449 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3450 * each single pixel it contains, and one U and one V value shared between both
3451 * pixels.
3452 *
3453 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3454 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3455 * take the format into account when generating the read swizzles
3456 *
3457 * Reading the Y value is streightforward - just sample the texture. The hardware
3458 * takes care of filtering in the horizontal and vertical direction.
3459 *
3460 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3461 * because that would mix the U and V values of one pixel or two adjacent pixels.
3462 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3463 * regardless of the filtering setting. Vertical filtering works automatically
3464 * though - the U and V values of two rows are mixed nicely.
3465 *
3466 * Appart of avoiding filtering issues, the code has to know which value it just
3467 * read, and where it can find the other one. To determine this, it checks if
3468 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3469 *
3470 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3471 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3472 *
3473 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3474 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3475 * in an unfiltered situation. Finding the luminance on the other hand requires
3476 * finding out if it is an odd or even pixel. The real drawback of this approach
3477 * is filtering. This would have to be emulated completely in the shader, reading
3478 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3479 * vertically. Beyond that it would require adjustments to the texture handling
3480 * code to deal with the width scaling
3481 */
3482 shader_addline(&buffer, "!!ARBfp1.0\n");
3483 shader_addline(&buffer, "TEMP luminance;\n");
3484 shader_addline(&buffer, "TEMP temp;\n");
3485 shader_addline(&buffer, "TEMP chroma;\n");
3486 shader_addline(&buffer, "TEMP texcrd;\n");
3487 shader_addline(&buffer, "TEMP texcrd2;\n");
3488 shader_addline(&buffer, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3489 shader_addline(&buffer, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3490 shader_addline(&buffer, "PARAM size = program.local[0];\n");
3491
3492 if(fmt == WINED3DFMT_UYVY || fmt ==WINED3DFMT_YUY2) {
3493 if(gen_planar_yuv_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3494 return 0;
3495 }
3496 } else {
3497 if(gen_yv12_read(&buffer, fmt, textype, &luminance_component) == FALSE) {
3498 return 0;
3499 }
3500 }
3501
3502 /* Calculate the final result. Formula is taken from
3503 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3504 * ranges from -0.5 to 0.5
3505 */
3506 shader_addline(&buffer, "SUB chroma.rg, chroma, coef.y;\n");
3507
3508 shader_addline(&buffer, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component);
3509 shader_addline(&buffer, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component);
3510 shader_addline(&buffer, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3511 shader_addline(&buffer, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component);
3512 shader_addline(&buffer, "END\n");
3513
3514 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(buffer.buffer), buffer.buffer));
3515
3516 if (glGetError() == GL_INVALID_OPERATION) {
3517 GLint pos;
3518 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &pos);
3519 FIXME("Fragment program error at position %d: %s\n", pos,
3520 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
3521 }
3522
3523 if(fmt == WINED3DFMT_YUY2) {
3524 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3525 priv->yuy2_rect_shader = shader;
3526 } else {
3527 priv->yuy2_2d_shader = shader;
3528 }
3529 } else if(fmt == WINED3DFMT_UYVY) {
3530 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3531 priv->uyvy_rect_shader = shader;
3532 } else {
3533 priv->uyvy_2d_shader = shader;
3534 }
3535 } else {
3536 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3537 priv->yv12_rect_shader = shader;
3538 } else {
3539 priv->yv12_2d_shader = shader;
3540 }
3541 }
3542 return shader;
3543 }
3544
3545 static HRESULT arbfp_blit_set(IWineD3DDevice *iface, WINED3DFORMAT fmt, GLenum textype, UINT width, UINT height) {
3546 GLenum shader;
3547 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3548 float size[4] = {width, height, 1, 1};
3549 struct arbfp_blit_priv *priv = (struct arbfp_blit_priv *) device->blit_priv;
3550 const GlPixelFormatDesc *glDesc;
3551
3552 getFormatDescEntry(fmt, &GLINFO_LOCATION, &glDesc);
3553
3554 if(glDesc->conversion_group != WINED3DFMT_YUY2 && glDesc->conversion_group != WINED3DFMT_UYVY &&
3555 glDesc->conversion_group != WINED3DFMT_YV12) {
3556 ERR("Format: %s\n", debug_d3dformat(glDesc->conversion_group));
3557 /* Don't bother setting up a shader for unconverted formats */
3558 glEnable(textype);
3559 checkGLcall("glEnable(textype)");
3560 return WINED3D_OK;
3561 }
3562
3563 if(glDesc->conversion_group == WINED3DFMT_YUY2) {
3564 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3565 shader = priv->yuy2_rect_shader;
3566 } else {
3567 shader = priv->yuy2_2d_shader;
3568 }
3569 } else if(glDesc->conversion_group == WINED3DFMT_UYVY) {
3570 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3571 shader = priv->uyvy_rect_shader;
3572 } else {
3573 shader = priv->uyvy_2d_shader;
3574 }
3575 } else {
3576 if(textype == GL_TEXTURE_RECTANGLE_ARB) {
3577 shader = priv->yv12_rect_shader;
3578 } else {
3579 shader = priv->yv12_2d_shader;
3580 }
3581 }
3582
3583 if(!shader) {
3584 shader = gen_yuv_shader(device, glDesc->conversion_group, textype);
3585 }
3586
3587 glEnable(GL_FRAGMENT_PROGRAM_ARB);
3588 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3589 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader));
3590 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3591 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0, size));
3592 checkGLcall("glProgramLocalParameter4fvARB");
3593
3594 return WINED3D_OK;
3595 }
3596
3597 static void arbfp_blit_unset(IWineD3DDevice *iface) {
3598 IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) iface;
3599 glDisable(GL_FRAGMENT_PROGRAM_ARB);
3600 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3601 glDisable(GL_TEXTURE_2D);
3602 checkGLcall("glDisable(GL_TEXTURE_2D)");
3603 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP)) {
3604 glDisable(GL_TEXTURE_CUBE_MAP_ARB);
3605 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3606 }
3607 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
3608 glDisable(GL_TEXTURE_RECTANGLE_ARB);
3609 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3610 }
3611 }
3612
3613 static BOOL arbfp_blit_conv_supported(WINED3DFORMAT fmt) {
3614 TRACE("Checking blit format support for format %s:", debug_d3dformat(fmt));
3615 switch(fmt) {
3616 case WINED3DFMT_YUY2:
3617 case WINED3DFMT_UYVY:
3618 case WINED3DFMT_YV12:
3619 TRACE("[OK]\n");
3620 return TRUE;
3621 default:
3622 TRACE("[FAILED]\n");
3623 return FALSE;
3624 }
3625 }
3626
3627 const struct blit_shader arbfp_blit = {
3628 arbfp_blit_alloc,
3629 arbfp_blit_free,
3630 arbfp_blit_set,
3631 arbfp_blit_unset,
3632 arbfp_blit_conv_supported
3633 };
3634
3635 #undef GLINFO_LOCATION
useful hacks and other patches not (yet) suitable for mainline